1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
2995
2996
2997
2998
2999
3000
3001
3002
3003
3004
3005
3006
3007
3008
3009
3010
3011
3012
3013
3014
3015
3016
3017
3018
3019
3020
3021
3022
3023
3024
3025
3026
3027
3028
3029
3030
3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
3053
3054
3055
3056
3057
3058
3059
3060
3061
3062
3063
3064
3065
3066
3067
3068
3069
3070
3071
3072
3073
3074
3075
3076
3077
3078
3079
3080
3081
3082
3083
3084
3085
3086
3087
3088
3089
3090
3091
3092
3093
3094
3095
3096
3097
3098
3099
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113
3114
3115
3116
3117
3118
3119
3120
3121
3122
3123
3124
3125
3126
3127
3128
3129
3130
3131
3132
3133
3134
3135
3136
3137
3138
3139
3140
3141
3142
3143
3144
3145
3146
3147
3148
3149
3150
3151
3152
3153
3154
3155
3156
3157
3158
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
3189
3190
3191
3192
3193
3194
3195
3196
3197
3198
3199
3200
3201
3202
3203
3204
3205
3206
3207
3208
3209
3210
3211
3212
3213
3214
3215
3216
3217
3218
3219
3220
3221
3222
3223
3224
3225
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247
3248
3249
3250
3251
3252
3253
3254
3255
3256
3257
3258
3259
3260
3261
3262
3263
3264
3265
3266
3267
3268
3269
3270
3271
3272
3273
3274
3275
3276
3277
3278
3279
3280
3281
3282
3283
3284
3285
3286
3287
3288
3289
3290
3291
3292
3293
3294
3295
3296
3297
3298
3299
3300
3301
3302
3303
3304
3305
3306
3307
3308
3309
3310
3311
3312
3313
3314
3315
3316
3317
3318
3319
3320
3321
3322
3323
3324
3325
3326
3327
3328
3329
3330
3331
3332
3333
3334
3335
3336
3337
3338
3339
3340
3341
3342
3343
3344
3345
3346
3347
3348
3349
3350
3351
3352
3353
3354
3355
3356
3357
3358
3359
3360
3361
3362
3363
3364
3365
3366
3367
3368
3369
3370
3371
3372
3373
3374
3375
3376
3377
3378
3379
3380
3381
3382
3383
3384
3385
3386
3387
3388
3389
3390
3391
3392
3393
3394
3395
3396
3397
3398
3399
3400
3401
3402
3403
3404
3405
3406
3407
3408
3409
3410
3411
3412
3413
3414
3415
3416
3417
3418
3419
3420
3421
3422
3423
3424
3425
3426
3427
3428
3429
3430
3431
3432
3433
3434
3435
3436
3437
3438
3439
3440
3441
3442
3443
3444
3445
3446
3447
3448
3449
3450
3451
3452
3453
3454
3455
3456
3457
3458
3459
3460
3461
3462
3463
3464
3465
3466
3467
3468
3469
3470
3471
3472
3473
3474
3475
3476
3477
3478
3479
3480
3481
3482
3483
3484
3485
3486
3487
3488
3489
3490
3491
3492
3493
3494
3495
3496
3497
3498
3499
3500
3501
3502
3503
3504
3505
3506
3507
3508
3509
3510
3511
3512
3513
3514
3515
3516
3517
3518
3519
3520
3521
3522
3523
3524
3525
3526
3527
3528
3529
3530
3531
3532
3533
3534
3535
3536
3537
3538
3539
3540
3541
3542
3543
3544
3545
3546
3547
3548
3549
3550
3551
3552
3553
3554
3555
3556
3557
3558
3559
3560
3561
3562
3563
3564
3565
3566
3567
3568
3569
3570
3571
3572
3573
3574
3575
3576
3577
3578
3579
3580
3581
3582
3583
3584
3585
3586
3587
3588
3589
3590
3591
3592
3593
3594
3595
3596
3597
3598
3599
3600
3601
3602
3603
3604
3605
3606
3607
3608
3609
3610
3611
3612
3613
3614
3615
3616
3617
3618
3619
3620
3621
3622
3623
3624
3625
3626
3627
3628
3629
3630
3631
3632
3633
3634
3635
3636
3637
3638
3639
3640
3641
3642
3643
3644
3645
3646
3647
3648
3649
3650
3651
3652
3653
3654
3655
3656
3657
3658
3659
3660
3661
3662
3663
3664
3665
3666
3667
3668
3669
3670
3671
3672
3673
3674
3675
3676
3677
3678
3679
3680
3681
3682
3683
3684
3685
3686
3687
3688
3689
3690
3691
3692
3693
3694
3695
3696
3697
3698
3699
3700
3701
3702
3703
3704
3705
3706
3707
3708
3709
3710
3711
3712
3713
3714
3715
3716
3717
3718
3719
3720
3721
3722
3723
3724
3725
3726
3727
3728
3729
3730
3731
3732
3733
3734
3735
3736
3737
3738
3739
3740
3741
3742
3743
3744
3745
3746
3747
3748
3749
3750
3751
3752
3753
3754
3755
3756
3757
3758
3759
3760
3761
3762
3763
3764
3765
3766
3767
3768
3769
3770
3771
3772
3773
3774
3775
3776
3777
3778
3779
3780
3781
3782
3783
3784
3785
3786
3787
3788
3789
3790
3791
3792
3793
3794
3795
3796
3797
3798
3799
3800
3801
3802
3803
3804
3805
3806
3807
3808
3809
3810
3811
3812
3813
3814
3815
3816
3817
3818
3819
3820
3821
3822
3823
3824
3825
3826
3827
3828
3829
3830
3831
3832
3833
3834
3835
3836
3837
3838
3839
3840
3841
3842
3843
3844
3845
3846
3847
3848
3849
3850
3851
3852
3853
3854
3855
3856
3857
3858
3859
3860
3861
3862
3863
3864
3865
3866
3867
3868
3869
3870
3871
3872
3873
3874
3875
3876
3877
3878
3879
3880
3881
3882
3883
3884
3885
3886
3887
3888
3889
3890
3891
3892
3893
3894
3895
3896
3897
3898
3899
3900
3901
3902
3903
3904
3905
3906
3907
3908
3909
3910
3911
3912
3913
3914
3915
3916
3917
3918
3919
3920
3921
3922
3923
3924
3925
3926
3927
3928
3929
3930
3931
3932
3933
3934
3935
3936
3937
3938
3939
3940
3941
3942
3943
3944
3945
3946
3947
3948
3949
3950
3951
3952
3953
3954
3955
3956
3957
3958
3959
3960
3961
3962
3963
3964
3965
3966
3967
3968
3969
3970
3971
3972
3973
3974
3975
3976
3977
3978
3979
3980
3981
3982
3983
3984
3985
3986
3987
3988
3989
3990
3991
3992
3993
3994
3995
3996
3997
3998
3999
4000
4001
4002
4003
4004
4005
4006
4007
4008
4009
4010
4011
4012
4013
4014
4015
4016
4017
4018
4019
4020
4021
4022
4023
4024
4025
4026
4027
4028
4029
4030
4031
4032
4033
4034
4035
4036
4037
4038
4039
4040
|
/*
* GK20A memory management
*
* Copyright (c) 2011-2016, NVIDIA CORPORATION. All rights reserved.
*
* This program is free software; you can redistribute it and/or modify it
* under the terms and conditions of the GNU General Public License,
* version 2, as published by the Free Software Foundation.
*
* This program is distributed in the hope it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
* more details.
*
* You should have received a copy of the GNU General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#include <linux/delay.h>
#include <linux/highmem.h>
#include <linux/log2.h>
#include <linux/nvhost.h>
#include <linux/pm_runtime.h>
#include <linux/scatterlist.h>
#include <linux/nvmap.h>
#include <linux/tegra-soc.h>
#include <linux/vmalloc.h>
#include <linux/dma-buf.h>
#include <linux/lcm.h>
#include <uapi/linux/nvgpu.h>
#include <trace/events/gk20a.h>
#include "gk20a.h"
#include "mm_gk20a.h"
#include "fence_gk20a.h"
#include "hw_gmmu_gk20a.h"
#include "hw_fb_gk20a.h"
#include "hw_bus_gk20a.h"
#include "hw_ram_gk20a.h"
#include "hw_mc_gk20a.h"
#include "hw_flush_gk20a.h"
#include "hw_ltc_gk20a.h"
#include "kind_gk20a.h"
#include "semaphore_gk20a.h"
int gk20a_mem_begin(struct gk20a *g, struct mem_desc *mem)
{
void *cpu_va;
if (WARN_ON(mem->cpu_va)) {
gk20a_warn(dev_from_gk20a(g), "nested %s", __func__);
return -EBUSY;
}
cpu_va = vmap(mem->pages,
PAGE_ALIGN(mem->size) >> PAGE_SHIFT,
0, pgprot_writecombine(PAGE_KERNEL));
if (WARN_ON(!cpu_va))
return -ENOMEM;
mem->cpu_va = cpu_va;
return 0;
}
void gk20a_mem_end(struct gk20a *g, struct mem_desc *mem)
{
vunmap(mem->cpu_va);
mem->cpu_va = NULL;
}
u32 gk20a_mem_rd32(struct gk20a *g, struct mem_desc *mem, u32 w)
{
u32 *ptr = mem->cpu_va;
u32 data;
WARN_ON(!ptr);
data = ptr[w];
#ifdef CONFIG_TEGRA_SIMULATION_PLATFORM
gk20a_dbg(gpu_dbg_mem, " %p = 0x%x", ptr + w, data);
#endif
return data;
}
u32 gk20a_mem_rd(struct gk20a *g, struct mem_desc *mem, u32 offset)
{
WARN_ON(offset & 3);
return gk20a_mem_rd32(g, mem, offset / sizeof(u32));
}
void gk20a_mem_rd_n(struct gk20a *g, struct mem_desc *mem,
u32 offset, void *dest, u32 size)
{
u32 i;
u32 *dest_u32 = dest;
WARN_ON(offset & 3);
WARN_ON(size & 3);
offset /= sizeof(u32);
size /= sizeof(u32);
for (i = 0; i < size; i++)
dest_u32[i] = gk20a_mem_rd32(g, mem, offset + i);
}
void gk20a_mem_wr32(struct gk20a *g, struct mem_desc *mem, u32 w, u32 data)
{
u32 *ptr = mem->cpu_va;
WARN_ON(!ptr);
#ifdef CONFIG_TEGRA_SIMULATION_PLATFORM
gk20a_dbg(gpu_dbg_mem, " %p = 0x%x", ptr + w, data);
#endif
ptr[w] = data;
}
void gk20a_mem_wr(struct gk20a *g, struct mem_desc *mem, u32 offset, u32 data)
{
WARN_ON(offset & 3);
gk20a_mem_wr32(g, mem, offset / sizeof(u32), data);
}
void gk20a_mem_wr_n(struct gk20a *g, struct mem_desc *mem, u32 offset,
void *src, u32 size)
{
u32 i;
u32 *src_u32 = src;
WARN_ON(offset & 3);
WARN_ON(size & 3);
offset /= sizeof(u32);
size /= sizeof(u32);
for (i = 0; i < size; i++)
gk20a_mem_wr32(g, mem, offset + i, src_u32[i]);
}
void gk20a_memset(struct gk20a *g, struct mem_desc *mem, u32 offset,
u32 value, u32 size)
{
u32 i;
WARN_ON(offset & 3);
WARN_ON(size & 3);
offset /= sizeof(u32);
size /= sizeof(u32);
for (i = 0; i < size; i++)
gk20a_mem_wr32(g, mem, offset + i, value);
}
/*
* GPU mapping life cycle
* ======================
*
* Kernel mappings
* ---------------
*
* Kernel mappings are created through vm.map(..., false):
*
* - Mappings to the same allocations are reused and refcounted.
* - This path does not support deferred unmapping (i.e. kernel must wait for
* all hw operations on the buffer to complete before unmapping).
* - References to dmabuf are owned and managed by the (kernel) clients of
* the gk20a_vm layer.
*
*
* User space mappings
* -------------------
*
* User space mappings are created through as.map_buffer -> vm.map(..., true):
*
* - Mappings to the same allocations are reused and refcounted.
* - This path supports deferred unmapping (i.e. we delay the actual unmapping
* until all hw operations have completed).
* - References to dmabuf are owned and managed by the vm_gk20a
* layer itself. vm.map acquires these refs, and sets
* mapped_buffer->own_mem_ref to record that we must release the refs when we
* actually unmap.
*
*/
static inline int vm_aspace_id(struct vm_gk20a *vm)
{
/* -1 is bar1 or pmu, etc. */
return vm->as_share ? vm->as_share->id : -1;
}
static inline u32 hi32(u64 f)
{
return (u32)(f >> 32);
}
static inline u32 lo32(u64 f)
{
return (u32)(f & 0xffffffff);
}
static struct mapped_buffer_node *find_mapped_buffer_locked(
struct rb_root *root, u64 addr);
static struct mapped_buffer_node *find_mapped_buffer_reverse_locked(
struct rb_root *root, struct dma_buf *dmabuf,
u32 kind);
static int update_gmmu_ptes_locked(struct vm_gk20a *vm,
enum gmmu_pgsz_gk20a pgsz_idx,
struct sg_table *sgt, u64 buffer_offset,
u64 first_vaddr, u64 last_vaddr,
u8 kind_v, u32 ctag_offset, bool cacheable,
bool umapped_pte, int rw_flag,
bool sparse,
bool priv);
static int __must_check gk20a_init_system_vm(struct mm_gk20a *mm);
static int __must_check gk20a_init_bar1_vm(struct mm_gk20a *mm);
static int __must_check gk20a_init_hwpm(struct mm_gk20a *mm);
static int __must_check gk20a_init_cde_vm(struct mm_gk20a *mm);
struct gk20a_dmabuf_priv {
struct mutex lock;
struct gk20a_comptag_allocator *comptag_allocator;
struct gk20a_comptags comptags;
struct dma_buf_attachment *attach;
struct sg_table *sgt;
int pin_count;
struct list_head states;
u64 buffer_id;
};
static void gk20a_vm_remove_support_nofree(struct vm_gk20a *vm);
static int gk20a_comptaglines_alloc(struct gk20a_comptag_allocator *allocator,
u32 *offset, u32 len)
{
unsigned long addr;
int err = 0;
mutex_lock(&allocator->lock);
addr = bitmap_find_next_zero_area(allocator->bitmap, allocator->size,
0, len, 0);
if (addr < allocator->size) {
/* number zero is reserved; bitmap base is 1 */
*offset = 1 + addr;
bitmap_set(allocator->bitmap, addr, len);
} else {
err = -ENOMEM;
}
mutex_unlock(&allocator->lock);
return err;
}
static void gk20a_comptaglines_free(struct gk20a_comptag_allocator *allocator,
u32 offset, u32 len)
{
/* number zero is reserved; bitmap base is 1 */
u32 addr = offset - 1;
WARN_ON(offset == 0);
WARN_ON(addr > allocator->size);
WARN_ON(addr + len > allocator->size);
mutex_lock(&allocator->lock);
bitmap_clear(allocator->bitmap, addr, len);
mutex_unlock(&allocator->lock);
}
static void gk20a_mm_delete_priv(void *_priv)
{
struct gk20a_buffer_state *s, *s_tmp;
struct gk20a_dmabuf_priv *priv = _priv;
if (!priv)
return;
if (priv->comptags.lines) {
BUG_ON(!priv->comptag_allocator);
gk20a_comptaglines_free(priv->comptag_allocator,
priv->comptags.offset,
priv->comptags.allocated_lines);
}
/* Free buffer states */
list_for_each_entry_safe(s, s_tmp, &priv->states, list) {
gk20a_fence_put(s->fence);
list_del(&s->list);
kfree(s);
}
kfree(priv);
}
struct sg_table *gk20a_mm_pin(struct device *dev, struct dma_buf *dmabuf)
{
struct gk20a_dmabuf_priv *priv;
priv = dma_buf_get_drvdata(dmabuf, dev);
if (WARN_ON(!priv))
return ERR_PTR(-EINVAL);
mutex_lock(&priv->lock);
if (priv->pin_count == 0) {
priv->attach = dma_buf_attach(dmabuf, dev);
if (IS_ERR(priv->attach)) {
mutex_unlock(&priv->lock);
return (struct sg_table *)priv->attach;
}
priv->sgt = dma_buf_map_attachment(priv->attach,
DMA_BIDIRECTIONAL);
if (IS_ERR(priv->sgt)) {
dma_buf_detach(dmabuf, priv->attach);
mutex_unlock(&priv->lock);
return priv->sgt;
}
}
priv->pin_count++;
mutex_unlock(&priv->lock);
return priv->sgt;
}
void gk20a_mm_unpin(struct device *dev, struct dma_buf *dmabuf,
struct sg_table *sgt)
{
struct gk20a_dmabuf_priv *priv = dma_buf_get_drvdata(dmabuf, dev);
dma_addr_t dma_addr;
if (IS_ERR(priv) || !priv)
return;
mutex_lock(&priv->lock);
WARN_ON(priv->sgt != sgt);
priv->pin_count--;
WARN_ON(priv->pin_count < 0);
dma_addr = sg_dma_address(priv->sgt->sgl);
if (priv->pin_count == 0) {
dma_buf_unmap_attachment(priv->attach, priv->sgt,
DMA_BIDIRECTIONAL);
dma_buf_detach(dmabuf, priv->attach);
}
mutex_unlock(&priv->lock);
}
void gk20a_get_comptags(struct device *dev, struct dma_buf *dmabuf,
struct gk20a_comptags *comptags)
{
struct gk20a_dmabuf_priv *priv = dma_buf_get_drvdata(dmabuf, dev);
if (!comptags)
return;
if (!priv) {
memset(comptags, 0, sizeof(*comptags));
return;
}
*comptags = priv->comptags;
}
static int gk20a_alloc_comptags(struct gk20a *g,
struct device *dev,
struct dma_buf *dmabuf,
struct gk20a_comptag_allocator *allocator,
u32 lines, bool user_mappable,
u64 *ctag_map_win_size,
u32 *ctag_map_win_ctagline)
{
struct gk20a_dmabuf_priv *priv = dma_buf_get_drvdata(dmabuf, dev);
u32 ctaglines_allocsize;
u32 ctagline_align;
u32 offset;
u32 alignment_lines;
const u32 aggregate_cacheline_sz =
g->gr.cacheline_size * g->gr.slices_per_ltc *
g->ltc_count;
const u32 small_pgsz = 4096;
int err;
if (!priv)
return -ENOSYS;
if (!lines)
return -EINVAL;
if (!user_mappable) {
ctaglines_allocsize = lines;
ctagline_align = 1;
} else {
/*
* For security, align the allocation on a page, and reserve
* whole pages. Unfortunately, we cannot ask the allocator to
* align here, since compbits per cacheline is not always a
* power of two. So, we just have to allocate enough extra that
* we're guaranteed to find a ctagline inside the allocation so
* that: 1) it is the first ctagline in a cacheline that starts
* at a page boundary, and 2) we can add enough overallocation
* that the ctaglines of the succeeding allocation are on
* different page than ours.
*/
ctagline_align =
(lcm(aggregate_cacheline_sz, small_pgsz) /
aggregate_cacheline_sz) *
g->gr.comptags_per_cacheline;
ctaglines_allocsize =
/* for alignment */
ctagline_align +
/* lines rounded up to cachelines */
DIV_ROUND_UP(lines, g->gr.comptags_per_cacheline) *
g->gr.comptags_per_cacheline +
/* trail-padding */
DIV_ROUND_UP(aggregate_cacheline_sz, small_pgsz) *
g->gr.comptags_per_cacheline;
if (ctaglines_allocsize < lines)
return -EINVAL; /* integer overflow */
}
/* store the allocator so we can use it when we free the ctags */
priv->comptag_allocator = allocator;
err = gk20a_comptaglines_alloc(allocator, &offset,
ctaglines_allocsize);
if (err)
return err;
/*
* offset needs to be at the start of a page/cacheline boundary;
* prune the preceding ctaglines that were allocated for alignment.
*/
alignment_lines =
DIV_ROUND_UP(offset, ctagline_align) * ctagline_align - offset;
if (alignment_lines) {
gk20a_comptaglines_free(allocator, offset, alignment_lines);
offset += alignment_lines;
ctaglines_allocsize -= alignment_lines;
}
/*
* check if we can prune the trailing, too; we just need to reserve
* whole pages and ctagcachelines.
*/
if (user_mappable) {
u32 needed_cachelines =
DIV_ROUND_UP(lines, g->gr.comptags_per_cacheline);
u32 needed_bytes = round_up(needed_cachelines *
aggregate_cacheline_sz,
small_pgsz);
u32 first_unneeded_cacheline =
DIV_ROUND_UP(needed_bytes, aggregate_cacheline_sz);
u32 needed_ctaglines = first_unneeded_cacheline *
g->gr.comptags_per_cacheline;
u64 win_size;
if (needed_ctaglines < ctaglines_allocsize) {
gk20a_comptaglines_free(allocator,
offset + needed_ctaglines,
ctaglines_allocsize - needed_ctaglines);
ctaglines_allocsize = needed_ctaglines;
}
*ctag_map_win_ctagline = offset;
win_size =
DIV_ROUND_UP(lines, g->gr.comptags_per_cacheline) *
aggregate_cacheline_sz;
*ctag_map_win_size = round_up(win_size, small_pgsz);
}
priv->comptags.offset = offset;
priv->comptags.lines = lines;
priv->comptags.allocated_lines = ctaglines_allocsize;
priv->comptags.user_mappable = user_mappable;
return 0;
}
static int gk20a_init_mm_reset_enable_hw(struct gk20a *g)
{
gk20a_dbg_fn("");
if (g->ops.fb.reset)
g->ops.fb.reset(g);
if (g->ops.clock_gating.slcg_fb_load_gating_prod)
g->ops.clock_gating.slcg_fb_load_gating_prod(g,
g->slcg_enabled);
if (g->ops.clock_gating.slcg_ltc_load_gating_prod)
g->ops.clock_gating.slcg_ltc_load_gating_prod(g,
g->slcg_enabled);
if (g->ops.clock_gating.blcg_fb_load_gating_prod)
g->ops.clock_gating.blcg_fb_load_gating_prod(g,
g->blcg_enabled);
if (g->ops.clock_gating.blcg_ltc_load_gating_prod)
g->ops.clock_gating.blcg_ltc_load_gating_prod(g,
g->blcg_enabled);
if (g->ops.fb.init_fs_state)
g->ops.fb.init_fs_state(g);
return 0;
}
void gk20a_remove_vm(struct vm_gk20a *vm, struct mem_desc *inst_block)
{
struct gk20a *g = vm->mm->g;
gk20a_dbg_fn("");
gk20a_free_inst_block(g, inst_block);
gk20a_vm_remove_support_nofree(vm);
}
static void gk20a_remove_mm_support(struct mm_gk20a *mm)
{
struct gk20a *g = gk20a_from_mm(mm);
if (g->ops.mm.remove_bar2_vm)
g->ops.mm.remove_bar2_vm(g);
gk20a_remove_vm(&mm->bar1.vm, &mm->bar1.inst_block);
gk20a_remove_vm(&mm->pmu.vm, &mm->pmu.inst_block);
gk20a_free_inst_block(gk20a_from_mm(mm), &mm->hwpm.inst_block);
gk20a_vm_remove_support_nofree(&mm->cde.vm);
}
static int gk20a_alloc_sysmem_flush(struct gk20a *g)
{
return gk20a_gmmu_alloc(g, SZ_4K, &g->mm.sysmem_flush);
}
int gk20a_init_mm_setup_sw(struct gk20a *g)
{
struct mm_gk20a *mm = &g->mm;
int err;
gk20a_dbg_fn("");
if (mm->sw_ready) {
gk20a_dbg_fn("skip init");
return 0;
}
mm->g = g;
mutex_init(&mm->l2_op_lock);
/*TBD: make channel vm size configurable */
mm->channel.user_size = NV_MM_DEFAULT_USER_SIZE;
mm->channel.kernel_size = NV_MM_DEFAULT_KERNEL_SIZE;
gk20a_dbg_info("channel vm size: user %dMB kernel %dMB",
(int)(mm->channel.user_size >> 20),
(int)(mm->channel.kernel_size >> 20));
err = gk20a_alloc_sysmem_flush(g);
if (err)
return err;
err = gk20a_init_bar1_vm(mm);
if (err)
return err;
if (g->ops.mm.init_bar2_vm) {
err = g->ops.mm.init_bar2_vm(g);
if (err)
return err;
}
err = gk20a_init_system_vm(mm);
if (err)
return err;
err = gk20a_init_hwpm(mm);
if (err)
return err;
err = gk20a_init_cde_vm(mm);
if (err)
return err;
/* set vm_alloc_share op here as gk20a_as_alloc_share needs it */
g->ops.mm.vm_alloc_share = gk20a_vm_alloc_share;
mm->remove_support = gk20a_remove_mm_support;
mm->sw_ready = true;
gk20a_dbg_fn("done");
return 0;
}
/* make sure gk20a_init_mm_support is called before */
int gk20a_init_mm_setup_hw(struct gk20a *g)
{
struct mm_gk20a *mm = &g->mm;
struct mem_desc *inst_block = &mm->bar1.inst_block;
u64 inst_pa = gk20a_mm_inst_block_addr(g, inst_block);
int err;
gk20a_dbg_fn("");
g->ops.fb.set_mmu_page_size(g);
if (g->ops.fb.set_use_full_comp_tag_line)
mm->use_full_comp_tag_line =
g->ops.fb.set_use_full_comp_tag_line(g);
gk20a_writel(g, fb_niso_flush_sysmem_addr_r(),
g->ops.mm.get_iova_addr(g, g->mm.sysmem_flush.sgt->sgl, 0)
>> 8);
g->ops.mm.bar1_bind(g, inst_pa);
if (g->ops.mm.init_bar2_mm_hw_setup) {
err = g->ops.mm.init_bar2_mm_hw_setup(g);
if (err)
return err;
}
if (gk20a_mm_fb_flush(g) || gk20a_mm_fb_flush(g))
return -EBUSY;
gk20a_dbg_fn("done");
return 0;
}
static int gk20a_mm_bar1_bind(struct gk20a *g, u64 bar1_iova)
{
u64 inst_pa = (u32)(bar1_iova >> bar1_instance_block_shift_gk20a());
gk20a_dbg_info("bar1 inst block ptr: 0x%08x", (u32)inst_pa);
gk20a_writel(g, bus_bar1_block_r(),
(g->mm.vidmem_is_vidmem ?
bus_bar1_block_target_sys_mem_ncoh_f() :
bus_bar1_block_target_vid_mem_f()) |
bus_bar1_block_mode_virtual_f() |
bus_bar1_block_ptr_f(inst_pa));
return 0;
}
int gk20a_init_mm_support(struct gk20a *g)
{
u32 err;
err = gk20a_init_mm_reset_enable_hw(g);
if (err)
return err;
err = gk20a_init_mm_setup_sw(g);
if (err)
return err;
if (g->ops.mm.init_mm_setup_hw)
err = g->ops.mm.init_mm_setup_hw(g);
return err;
}
static int alloc_gmmu_phys_pages(struct vm_gk20a *vm, u32 order,
struct gk20a_mm_entry *entry)
{
u32 num_pages = 1 << order;
u32 len = num_pages * PAGE_SIZE;
int err;
struct page *pages;
gk20a_dbg_fn("");
/* note: mem_desc slightly abused (wrt. alloc_gmmu_pages) */
pages = alloc_pages(GFP_KERNEL, order);
if (!pages) {
gk20a_dbg(gpu_dbg_pte, "alloc_pages failed");
goto err_out;
}
entry->mem.sgt = kzalloc(sizeof(*entry->mem.sgt), GFP_KERNEL);
if (!entry->mem.sgt) {
gk20a_dbg(gpu_dbg_pte, "cannot allocate sg table");
goto err_alloced;
}
err = sg_alloc_table(entry->mem.sgt, 1, GFP_KERNEL);
if (err) {
gk20a_dbg(gpu_dbg_pte, "sg_alloc_table failed");
goto err_sg_table;
}
sg_set_page(entry->mem.sgt->sgl, pages, len, 0);
entry->mem.cpu_va = page_address(pages);
memset(entry->mem.cpu_va, 0, len);
entry->mem.size = len;
FLUSH_CPU_DCACHE(entry->mem.cpu_va, sg_phys(entry->mem.sgt->sgl), len);
return 0;
err_sg_table:
kfree(entry->mem.sgt);
err_alloced:
__free_pages(pages, order);
err_out:
return -ENOMEM;
}
static void free_gmmu_phys_pages(struct vm_gk20a *vm,
struct gk20a_mm_entry *entry)
{
gk20a_dbg_fn("");
/* note: mem_desc slightly abused (wrt. free_gmmu_pages) */
free_pages((unsigned long)entry->mem.cpu_va, get_order(entry->mem.size));
entry->mem.cpu_va = NULL;
sg_free_table(entry->mem.sgt);
kfree(entry->mem.sgt);
entry->mem.sgt = NULL;
entry->mem.size = 0;
}
static int map_gmmu_phys_pages(struct gk20a_mm_entry *entry)
{
FLUSH_CPU_DCACHE(entry->mem.cpu_va,
sg_phys(entry->mem.sgt->sgl),
entry->mem.sgt->sgl->length);
return 0;
}
static void unmap_gmmu_phys_pages(struct gk20a_mm_entry *entry)
{
FLUSH_CPU_DCACHE(entry->mem.cpu_va,
sg_phys(entry->mem.sgt->sgl),
entry->mem.sgt->sgl->length);
}
static int alloc_gmmu_pages(struct vm_gk20a *vm, u32 order,
struct gk20a_mm_entry *entry)
{
struct device *d = dev_from_vm(vm);
struct gk20a *g = gk20a_from_vm(vm);
u32 num_pages = 1 << order;
u32 len = num_pages * PAGE_SIZE;
int err;
gk20a_dbg_fn("");
if (tegra_platform_is_linsim())
return alloc_gmmu_phys_pages(vm, order, entry);
/*
* On arm32 we're limited by vmalloc space, so we do not map pages by
* default.
*/
if (IS_ENABLED(CONFIG_ARM64))
err = gk20a_gmmu_alloc(g, len, &entry->mem);
else
err = gk20a_gmmu_alloc_attr(g, DMA_ATTR_NO_KERNEL_MAPPING,
len, &entry->mem);
if (err) {
gk20a_err(d, "memory allocation failed");
return -ENOMEM;
}
return 0;
}
void free_gmmu_pages(struct vm_gk20a *vm,
struct gk20a_mm_entry *entry)
{
struct gk20a *g = gk20a_from_vm(vm);
gk20a_dbg_fn("");
if (!entry->mem.size)
return;
if (tegra_platform_is_linsim()) {
free_gmmu_phys_pages(vm, entry);
return;
}
/*
* On arm32 we're limited by vmalloc space, so we do not map pages by
* default.
*/
if (IS_ENABLED(CONFIG_ARM64))
gk20a_gmmu_free(g, &entry->mem);
else
gk20a_gmmu_free_attr(g, DMA_ATTR_NO_KERNEL_MAPPING,
&entry->mem);
}
int map_gmmu_pages(struct gk20a_mm_entry *entry)
{
int count = PAGE_ALIGN(entry->mem.size) >> PAGE_SHIFT;
struct page **pages;
gk20a_dbg_fn("");
if (tegra_platform_is_linsim())
return map_gmmu_phys_pages(entry);
if (IS_ENABLED(CONFIG_ARM64)) {
FLUSH_CPU_DCACHE(entry->mem.cpu_va,
sg_phys(entry->mem.sgt->sgl),
entry->mem.size);
} else {
pages = entry->mem.pages;
entry->mem.cpu_va = vmap(pages, count, 0,
pgprot_writecombine(PAGE_KERNEL));
if (!entry->mem.cpu_va)
return -ENOMEM;
}
return 0;
}
void unmap_gmmu_pages(struct gk20a_mm_entry *entry)
{
gk20a_dbg_fn("");
if (tegra_platform_is_linsim()) {
unmap_gmmu_phys_pages(entry);
return;
}
if (IS_ENABLED(CONFIG_ARM64)) {
FLUSH_CPU_DCACHE(entry->mem.cpu_va,
sg_phys(entry->mem.sgt->sgl),
entry->mem.size);
} else {
vunmap(entry->mem.cpu_va);
entry->mem.cpu_va = NULL;
}
}
/* allocate a phys contig region big enough for a full
* sized gmmu page table for the given gmmu_page_size.
* the whole range is zeroed so it's "invalid"/will fault
*/
static int gk20a_zalloc_gmmu_page_table(struct vm_gk20a *vm,
enum gmmu_pgsz_gk20a pgsz_idx,
const struct gk20a_mmu_level *l,
struct gk20a_mm_entry *entry)
{
int err;
int order;
struct gk20a *g = gk20a_from_vm(vm);
gk20a_dbg_fn("");
/* allocate enough pages for the table */
order = l->hi_bit[pgsz_idx] - l->lo_bit[pgsz_idx] + 1;
order += ilog2(l->entry_size);
order -= PAGE_SHIFT;
order = max(0, order);
err = alloc_gmmu_pages(vm, order, entry);
gk20a_dbg(gpu_dbg_pte, "entry = 0x%p, addr=%08llx, size %d",
entry, g->ops.mm.get_iova_addr(g, entry->mem.sgt->sgl, 0),
order);
if (err)
return err;
entry->pgsz = pgsz_idx;
return err;
}
int gk20a_mm_pde_coverage_bit_count(struct vm_gk20a *vm)
{
return vm->mmu_levels[0].lo_bit[0];
}
/* given address range (inclusive) determine the pdes crossed */
void pde_range_from_vaddr_range(struct vm_gk20a *vm,
u64 addr_lo, u64 addr_hi,
u32 *pde_lo, u32 *pde_hi)
{
int pde_shift = gk20a_mm_pde_coverage_bit_count(vm);
*pde_lo = (u32)(addr_lo >> pde_shift);
*pde_hi = (u32)(addr_hi >> pde_shift);
gk20a_dbg(gpu_dbg_pte, "addr_lo=0x%llx addr_hi=0x%llx pde_ss=%d",
addr_lo, addr_hi, pde_shift);
gk20a_dbg(gpu_dbg_pte, "pde_lo=%d pde_hi=%d",
*pde_lo, *pde_hi);
}
static u32 pde_from_index(u32 i)
{
return i * gmmu_pde__size_v() / sizeof(u32);
}
static u32 pte_from_index(u32 i)
{
return i * gmmu_pte__size_v() / sizeof(u32);
}
u32 pte_index_from_vaddr(struct vm_gk20a *vm,
u64 addr, enum gmmu_pgsz_gk20a pgsz_idx)
{
u32 ret;
/* mask off pde part */
addr = addr & ((1ULL << gk20a_mm_pde_coverage_bit_count(vm)) - 1ULL);
/* shift over to get pte index. note assumption that pte index
* doesn't leak over into the high 32b */
ret = (u32)(addr >> ilog2(vm->gmmu_page_sizes[pgsz_idx]));
gk20a_dbg(gpu_dbg_pte, "addr=0x%llx pte_i=0x%x", addr, ret);
return ret;
}
static struct vm_reserved_va_node *addr_to_reservation(struct vm_gk20a *vm,
u64 addr)
{
struct vm_reserved_va_node *va_node;
list_for_each_entry(va_node, &vm->reserved_va_list, reserved_va_list)
if (addr >= va_node->vaddr_start &&
addr < (u64)va_node->vaddr_start + (u64)va_node->size)
return va_node;
return NULL;
}
int gk20a_vm_get_buffers(struct vm_gk20a *vm,
struct mapped_buffer_node ***mapped_buffers,
int *num_buffers)
{
struct mapped_buffer_node *mapped_buffer;
struct mapped_buffer_node **buffer_list;
struct rb_node *node;
int i = 0;
if (vm->userspace_managed) {
*mapped_buffers = NULL;
*num_buffers = 0;
return 0;
}
mutex_lock(&vm->update_gmmu_lock);
buffer_list = nvgpu_alloc(sizeof(*buffer_list) *
vm->num_user_mapped_buffers, true);
if (!buffer_list) {
mutex_unlock(&vm->update_gmmu_lock);
return -ENOMEM;
}
node = rb_first(&vm->mapped_buffers);
while (node) {
mapped_buffer =
container_of(node, struct mapped_buffer_node, node);
if (mapped_buffer->user_mapped) {
buffer_list[i] = mapped_buffer;
kref_get(&mapped_buffer->ref);
i++;
}
node = rb_next(&mapped_buffer->node);
}
BUG_ON(i != vm->num_user_mapped_buffers);
*num_buffers = vm->num_user_mapped_buffers;
*mapped_buffers = buffer_list;
mutex_unlock(&vm->update_gmmu_lock);
return 0;
}
static void gk20a_vm_unmap_locked_kref(struct kref *ref)
{
struct mapped_buffer_node *mapped_buffer =
container_of(ref, struct mapped_buffer_node, ref);
gk20a_vm_unmap_locked(mapped_buffer, mapped_buffer->vm->kref_put_batch);
}
void gk20a_vm_mapping_batch_start(struct vm_gk20a_mapping_batch *mapping_batch)
{
memset(mapping_batch, 0, sizeof(*mapping_batch));
mapping_batch->gpu_l2_flushed = false;
mapping_batch->need_tlb_invalidate = false;
}
void gk20a_vm_mapping_batch_finish_locked(
struct vm_gk20a *vm, struct vm_gk20a_mapping_batch *mapping_batch)
{
/* hanging kref_put batch pointer? */
WARN_ON(vm->kref_put_batch == mapping_batch);
if (mapping_batch->need_tlb_invalidate) {
struct gk20a *g = gk20a_from_vm(vm);
g->ops.mm.tlb_invalidate(vm);
}
}
void gk20a_vm_mapping_batch_finish(struct vm_gk20a *vm,
struct vm_gk20a_mapping_batch *mapping_batch)
{
mutex_lock(&vm->update_gmmu_lock);
gk20a_vm_mapping_batch_finish_locked(vm, mapping_batch);
mutex_unlock(&vm->update_gmmu_lock);
}
void gk20a_vm_put_buffers(struct vm_gk20a *vm,
struct mapped_buffer_node **mapped_buffers,
int num_buffers)
{
int i;
struct vm_gk20a_mapping_batch batch;
if (num_buffers == 0)
return;
mutex_lock(&vm->update_gmmu_lock);
gk20a_vm_mapping_batch_start(&batch);
vm->kref_put_batch = &batch;
for (i = 0; i < num_buffers; ++i)
kref_put(&mapped_buffers[i]->ref,
gk20a_vm_unmap_locked_kref);
vm->kref_put_batch = NULL;
gk20a_vm_mapping_batch_finish_locked(vm, &batch);
mutex_unlock(&vm->update_gmmu_lock);
nvgpu_free(mapped_buffers);
}
static void gk20a_vm_unmap_user(struct vm_gk20a *vm, u64 offset,
struct vm_gk20a_mapping_batch *batch)
{
struct device *d = dev_from_vm(vm);
int retries = 10000; /* 50 ms */
struct mapped_buffer_node *mapped_buffer;
mutex_lock(&vm->update_gmmu_lock);
mapped_buffer = find_mapped_buffer_locked(&vm->mapped_buffers, offset);
if (!mapped_buffer) {
mutex_unlock(&vm->update_gmmu_lock);
gk20a_err(d, "invalid addr to unmap 0x%llx", offset);
return;
}
if (mapped_buffer->flags & NVGPU_AS_MAP_BUFFER_FLAGS_FIXED_OFFSET) {
mutex_unlock(&vm->update_gmmu_lock);
while (retries >= 0 || !tegra_platform_is_silicon()) {
if (atomic_read(&mapped_buffer->ref.refcount) == 1)
break;
retries--;
udelay(5);
}
if (retries < 0 && tegra_platform_is_silicon())
gk20a_err(d, "sync-unmap failed on 0x%llx",
offset);
mutex_lock(&vm->update_gmmu_lock);
}
if (mapped_buffer->user_mapped == 0) {
mutex_unlock(&vm->update_gmmu_lock);
gk20a_err(d, "addr already unmapped from user 0x%llx", offset);
return;
}
mapped_buffer->user_mapped--;
if (mapped_buffer->user_mapped == 0)
vm->num_user_mapped_buffers--;
vm->kref_put_batch = batch;
kref_put(&mapped_buffer->ref, gk20a_vm_unmap_locked_kref);
vm->kref_put_batch = NULL;
mutex_unlock(&vm->update_gmmu_lock);
}
u64 gk20a_vm_alloc_va(struct vm_gk20a *vm,
u64 size,
enum gmmu_pgsz_gk20a gmmu_pgsz_idx)
{
struct gk20a_allocator *vma = &vm->vma[gmmu_pgsz_idx];
u64 offset;
u64 gmmu_page_size = vm->gmmu_page_sizes[gmmu_pgsz_idx];
if (gmmu_pgsz_idx >= gmmu_nr_page_sizes) {
dev_warn(dev_from_vm(vm),
"invalid page size requested in gk20a vm alloc");
return 0;
}
if ((gmmu_pgsz_idx == gmmu_page_size_big) && !vm->big_pages) {
dev_warn(dev_from_vm(vm),
"unsupportd page size requested");
return 0;
}
/* Be certain we round up to gmmu_page_size if needed */
size = (size + ((u64)gmmu_page_size - 1)) & ~((u64)gmmu_page_size - 1);
gk20a_dbg_info("size=0x%llx @ pgsz=%dKB", size,
vm->gmmu_page_sizes[gmmu_pgsz_idx]>>10);
offset = gk20a_balloc(vma, size);
if (!offset) {
gk20a_err(dev_from_vm(vm),
"%s oom: sz=0x%llx", vma->name, size);
return 0;
}
gk20a_dbg_fn("%s found addr: 0x%llx", vma->name, offset);
return offset;
}
int gk20a_vm_free_va(struct vm_gk20a *vm,
u64 offset, u64 size,
enum gmmu_pgsz_gk20a pgsz_idx)
{
struct gk20a_allocator *vma = &vm->vma[pgsz_idx];
gk20a_dbg_info("%s free addr=0x%llx, size=0x%llx",
vma->name, offset, size);
gk20a_bfree(vma, offset);
return 0;
}
static int insert_mapped_buffer(struct rb_root *root,
struct mapped_buffer_node *mapped_buffer)
{
struct rb_node **new_node = &(root->rb_node), *parent = NULL;
/* Figure out where to put new node */
while (*new_node) {
struct mapped_buffer_node *cmp_with =
container_of(*new_node, struct mapped_buffer_node,
node);
parent = *new_node;
if (cmp_with->addr > mapped_buffer->addr) /* u64 cmp */
new_node = &((*new_node)->rb_left);
else if (cmp_with->addr != mapped_buffer->addr) /* u64 cmp */
new_node = &((*new_node)->rb_right);
else
return -EINVAL; /* no fair dup'ing */
}
/* Add new node and rebalance tree. */
rb_link_node(&mapped_buffer->node, parent, new_node);
rb_insert_color(&mapped_buffer->node, root);
return 0;
}
static struct mapped_buffer_node *find_mapped_buffer_reverse_locked(
struct rb_root *root, struct dma_buf *dmabuf,
u32 kind)
{
struct rb_node *node = rb_first(root);
while (node) {
struct mapped_buffer_node *mapped_buffer =
container_of(node, struct mapped_buffer_node, node);
if (mapped_buffer->dmabuf == dmabuf &&
kind == mapped_buffer->kind)
return mapped_buffer;
node = rb_next(&mapped_buffer->node);
}
return NULL;
}
static struct mapped_buffer_node *find_mapped_buffer_locked(
struct rb_root *root, u64 addr)
{
struct rb_node *node = root->rb_node;
while (node) {
struct mapped_buffer_node *mapped_buffer =
container_of(node, struct mapped_buffer_node, node);
if (mapped_buffer->addr > addr) /* u64 cmp */
node = node->rb_left;
else if (mapped_buffer->addr != addr) /* u64 cmp */
node = node->rb_right;
else
return mapped_buffer;
}
return NULL;
}
static struct mapped_buffer_node *find_mapped_buffer_range_locked(
struct rb_root *root, u64 addr)
{
struct rb_node *node = root->rb_node;
while (node) {
struct mapped_buffer_node *m =
container_of(node, struct mapped_buffer_node, node);
if (m->addr <= addr && m->addr + m->size > addr)
return m;
else if (m->addr > addr) /* u64 cmp */
node = node->rb_left;
else
node = node->rb_right;
}
return NULL;
}
/* find the first mapped buffer with GPU VA less than addr */
static struct mapped_buffer_node *find_mapped_buffer_less_than_locked(
struct rb_root *root, u64 addr)
{
struct rb_node *node = root->rb_node;
struct mapped_buffer_node *ret = NULL;
while (node) {
struct mapped_buffer_node *mapped_buffer =
container_of(node, struct mapped_buffer_node, node);
if (mapped_buffer->addr >= addr)
node = node->rb_left;
else {
ret = mapped_buffer;
node = node->rb_right;
}
}
return ret;
}
#define BFR_ATTRS (sizeof(nvmap_bfr_param)/sizeof(nvmap_bfr_param[0]))
struct buffer_attrs {
struct sg_table *sgt;
u64 size;
u64 align;
u32 ctag_offset;
u32 ctag_lines;
u32 ctag_allocated_lines;
int pgsz_idx;
u8 kind_v;
u8 uc_kind_v;
bool ctag_user_mappable;
};
static void gmmu_select_page_size(struct vm_gk20a *vm,
struct buffer_attrs *bfr)
{
int i;
/* choose the biggest first (top->bottom) */
for (i = gmmu_page_size_kernel - 1; i >= 0; i--)
if (!((vm->gmmu_page_sizes[i] - 1) & bfr->align)) {
bfr->pgsz_idx = i;
break;
}
}
static int setup_buffer_kind_and_compression(struct vm_gk20a *vm,
u32 flags,
struct buffer_attrs *bfr,
enum gmmu_pgsz_gk20a pgsz_idx)
{
bool kind_compressible;
struct gk20a *g = gk20a_from_vm(vm);
struct device *d = dev_from_gk20a(g);
int ctag_granularity = g->ops.fb.compression_page_size(g);
if (unlikely(bfr->kind_v == gmmu_pte_kind_invalid_v()))
bfr->kind_v = gmmu_pte_kind_pitch_v();
if (unlikely(!gk20a_kind_is_supported(bfr->kind_v))) {
gk20a_err(d, "kind 0x%x not supported", bfr->kind_v);
return -EINVAL;
}
bfr->uc_kind_v = gmmu_pte_kind_invalid_v();
/* find a suitable uncompressed kind if it becomes necessary later */
kind_compressible = gk20a_kind_is_compressible(bfr->kind_v);
if (kind_compressible) {
bfr->uc_kind_v = gk20a_get_uncompressed_kind(bfr->kind_v);
if (unlikely(bfr->uc_kind_v == gmmu_pte_kind_invalid_v())) {
/* shouldn't happen, but it is worth cross-checking */
gk20a_err(d, "comptag kind 0x%x can't be"
" downgraded to uncompressed kind",
bfr->kind_v);
return -EINVAL;
}
}
/* comptags only supported for suitable kinds, 128KB pagesize */
if (kind_compressible &&
vm->gmmu_page_sizes[pgsz_idx] < g->ops.fb.compressible_page_size(g)) {
/*
gk20a_warn(d, "comptags specified"
" but pagesize being used doesn't support it");*/
/* it is safe to fall back to uncompressed as
functionality is not harmed */
bfr->kind_v = bfr->uc_kind_v;
kind_compressible = false;
}
if (kind_compressible)
bfr->ctag_lines = DIV_ROUND_UP_ULL(bfr->size, ctag_granularity);
else
bfr->ctag_lines = 0;
return 0;
}
static int validate_fixed_buffer(struct vm_gk20a *vm,
struct buffer_attrs *bfr,
u64 map_offset, u64 map_size,
struct vm_reserved_va_node **pva_node)
{
struct device *dev = dev_from_vm(vm);
struct vm_reserved_va_node *va_node;
struct mapped_buffer_node *buffer;
u64 map_end = map_offset + map_size;
/* can wrap around with insane map_size; zero is disallowed too */
if (map_end <= map_offset) {
gk20a_warn(dev, "fixed offset mapping with invalid map_size");
return -EINVAL;
}
if (map_offset & (vm->gmmu_page_sizes[bfr->pgsz_idx] - 1)) {
gk20a_err(dev, "map offset must be buffer page size aligned 0x%llx",
map_offset);
return -EINVAL;
}
/* Find the space reservation, but it's ok to have none for
* userspace-managed address spaces */
va_node = addr_to_reservation(vm, map_offset);
if (!va_node && !vm->userspace_managed) {
gk20a_warn(dev, "fixed offset mapping without space allocation");
return -EINVAL;
}
/* Mapped area should fit inside va, if there's one */
if (va_node && map_end > va_node->vaddr_start + va_node->size) {
gk20a_warn(dev, "fixed offset mapping size overflows va node");
return -EINVAL;
}
/* check that this mapping does not collide with existing
* mappings by checking the buffer with the highest GPU VA
* that is less than our buffer end */
buffer = find_mapped_buffer_less_than_locked(
&vm->mapped_buffers, map_offset + map_size);
if (buffer && buffer->addr + buffer->size > map_offset) {
gk20a_warn(dev, "overlapping buffer map requested");
return -EINVAL;
}
*pva_node = va_node;
return 0;
}
u64 gk20a_locked_gmmu_map(struct vm_gk20a *vm,
u64 map_offset,
struct sg_table *sgt,
u64 buffer_offset,
u64 size,
int pgsz_idx,
u8 kind_v,
u32 ctag_offset,
u32 flags,
int rw_flag,
bool clear_ctags,
bool sparse,
bool priv,
struct vm_gk20a_mapping_batch *batch)
{
int err = 0;
bool allocated = false;
struct device *d = dev_from_vm(vm);
struct gk20a *g = gk20a_from_vm(vm);
int ctag_granularity = g->ops.fb.compression_page_size(g);
u32 ctag_lines = DIV_ROUND_UP_ULL(size, ctag_granularity);
/* Allocate (or validate when map_offset != 0) the virtual address. */
if (!map_offset) {
map_offset = gk20a_vm_alloc_va(vm, size,
pgsz_idx);
if (!map_offset) {
gk20a_err(d, "failed to allocate va space");
err = -ENOMEM;
goto fail_alloc;
}
allocated = true;
}
gk20a_dbg(gpu_dbg_map,
"as=%d pgsz=%d "
"kind=0x%x flags=0x%x "
"ctags=%d start=%d gv=0x%x,%08x -> 0x%x,%08x -> 0x%x,%08x + %llx",
vm_aspace_id(vm), pgsz_idx,
kind_v, flags,
ctag_lines, ctag_offset,
hi32(map_offset), lo32(map_offset),
hi32((u64)sg_dma_address(sgt->sgl)),
lo32((u64)sg_dma_address(sgt->sgl)),
hi32((u64)sg_phys(sgt->sgl)),
lo32((u64)sg_phys(sgt->sgl)),
buffer_offset);
err = update_gmmu_ptes_locked(vm, pgsz_idx,
sgt,
buffer_offset,
map_offset, map_offset + size,
kind_v,
ctag_offset,
flags &
NVGPU_MAP_BUFFER_FLAGS_CACHEABLE_TRUE,
flags &
NVGPU_AS_MAP_BUFFER_FLAGS_UNMAPPED_PTE,
rw_flag,
sparse,
priv);
if (err) {
gk20a_err(d, "failed to update ptes on map");
goto fail_validate;
}
if (!batch)
g->ops.mm.tlb_invalidate(vm);
else
batch->need_tlb_invalidate = true;
return map_offset;
fail_validate:
if (allocated)
gk20a_vm_free_va(vm, map_offset, size, pgsz_idx);
fail_alloc:
gk20a_err(d, "%s: failed with err=%d\n", __func__, err);
return 0;
}
void gk20a_locked_gmmu_unmap(struct vm_gk20a *vm,
u64 vaddr,
u64 size,
int pgsz_idx,
bool va_allocated,
int rw_flag,
bool sparse,
struct vm_gk20a_mapping_batch *batch)
{
int err = 0;
struct gk20a *g = gk20a_from_vm(vm);
if (va_allocated) {
err = gk20a_vm_free_va(vm, vaddr, size, pgsz_idx);
if (err) {
dev_err(dev_from_vm(vm),
"failed to free va");
return;
}
}
/* unmap here needs to know the page size we assigned at mapping */
err = update_gmmu_ptes_locked(vm,
pgsz_idx,
NULL, /* n/a for unmap */
0,
vaddr,
vaddr + size,
0, 0, false /* n/a for unmap */,
false, rw_flag,
sparse, 0);
if (err)
dev_err(dev_from_vm(vm),
"failed to update gmmu ptes on unmap");
/* flush l2 so any dirty lines are written out *now*.
* also as we could potentially be switching this buffer
* from nonvolatile (l2 cacheable) to volatile (l2 non-cacheable) at
* some point in the future we need to invalidate l2. e.g. switching
* from a render buffer unmap (here) to later using the same memory
* for gmmu ptes. note the positioning of this relative to any smmu
* unmapping (below). */
if (!batch) {
gk20a_mm_l2_flush(g, true);
g->ops.mm.tlb_invalidate(vm);
} else {
if (!batch->gpu_l2_flushed) {
gk20a_mm_l2_flush(g, true);
batch->gpu_l2_flushed = true;
}
batch->need_tlb_invalidate = true;
}
}
static u64 gk20a_vm_map_duplicate_locked(struct vm_gk20a *vm,
struct dma_buf *dmabuf,
u64 offset_align,
u32 flags,
int kind,
struct sg_table **sgt,
bool user_mapped,
int rw_flag)
{
struct mapped_buffer_node *mapped_buffer = NULL;
mapped_buffer =
find_mapped_buffer_reverse_locked(&vm->mapped_buffers,
dmabuf, kind);
if (!mapped_buffer)
return 0;
if (mapped_buffer->flags != flags)
return 0;
if (flags & NVGPU_AS_MAP_BUFFER_FLAGS_FIXED_OFFSET &&
mapped_buffer->addr != offset_align)
return 0;
BUG_ON(mapped_buffer->vm != vm);
/* mark the buffer as used */
if (user_mapped) {
if (mapped_buffer->user_mapped == 0)
vm->num_user_mapped_buffers++;
mapped_buffer->user_mapped++;
/* If the mapping comes from user space, we own
* the handle ref. Since we reuse an
* existing mapping here, we need to give back those
* refs once in order not to leak.
*/
if (mapped_buffer->own_mem_ref)
dma_buf_put(mapped_buffer->dmabuf);
else
mapped_buffer->own_mem_ref = true;
}
kref_get(&mapped_buffer->ref);
gk20a_dbg(gpu_dbg_map,
"reusing as=%d pgsz=%d flags=0x%x ctags=%d "
"start=%d gv=0x%x,%08x -> 0x%x,%08x -> 0x%x,%08x "
"own_mem_ref=%d user_mapped=%d",
vm_aspace_id(vm), mapped_buffer->pgsz_idx,
mapped_buffer->flags,
mapped_buffer->ctag_lines,
mapped_buffer->ctag_offset,
hi32(mapped_buffer->addr), lo32(mapped_buffer->addr),
hi32((u64)sg_dma_address(mapped_buffer->sgt->sgl)),
lo32((u64)sg_dma_address(mapped_buffer->sgt->sgl)),
hi32((u64)sg_phys(mapped_buffer->sgt->sgl)),
lo32((u64)sg_phys(mapped_buffer->sgt->sgl)),
mapped_buffer->own_mem_ref, user_mapped);
if (sgt)
*sgt = mapped_buffer->sgt;
return mapped_buffer->addr;
}
u64 gk20a_vm_map(struct vm_gk20a *vm,
struct dma_buf *dmabuf,
u64 offset_align,
u32 flags /*NVGPU_AS_MAP_BUFFER_FLAGS_*/,
int kind,
struct sg_table **sgt,
bool user_mapped,
int rw_flag,
u64 buffer_offset,
u64 mapping_size,
struct vm_gk20a_mapping_batch *batch)
{
struct gk20a *g = gk20a_from_vm(vm);
struct gk20a_comptag_allocator *ctag_allocator = &g->gr.comp_tags;
struct device *d = dev_from_vm(vm);
struct mapped_buffer_node *mapped_buffer = NULL;
bool inserted = false, va_allocated = false;
u32 gmmu_page_size = 0;
u64 map_offset = 0;
int err = 0;
struct buffer_attrs bfr = {NULL};
struct gk20a_comptags comptags;
bool clear_ctags = false;
struct scatterlist *sgl;
u64 buf_addr;
u64 ctag_map_win_size = 0;
u32 ctag_map_win_ctagline = 0;
struct vm_reserved_va_node *va_node = NULL;
u32 ctag_offset;
if (user_mapped && vm->userspace_managed &&
!(flags & NVGPU_AS_MAP_BUFFER_FLAGS_FIXED_OFFSET)) {
gk20a_err(d,
"%s: non-fixed-offset mapping not available on userspace managed address spaces",
__func__);
return -EFAULT;
}
mutex_lock(&vm->update_gmmu_lock);
/* check if this buffer is already mapped */
if (!vm->userspace_managed) {
map_offset = gk20a_vm_map_duplicate_locked(
vm, dmabuf, offset_align,
flags, kind, sgt,
user_mapped, rw_flag);
if (map_offset) {
mutex_unlock(&vm->update_gmmu_lock);
return map_offset;
}
}
/* pin buffer to get phys/iovmm addr */
bfr.sgt = gk20a_mm_pin(d, dmabuf);
if (IS_ERR(bfr.sgt)) {
/* Falling back to physical is actually possible
* here in many cases if we use 4K phys pages in the
* gmmu. However we have some regions which require
* contig regions to work properly (either phys-contig
* or contig through smmu io_vaspace). Until we can
* track the difference between those two cases we have
* to fail the mapping when we run out of SMMU space.
*/
gk20a_warn(d, "oom allocating tracking buffer");
goto clean_up;
}
if (sgt)
*sgt = bfr.sgt;
bfr.kind_v = kind;
bfr.size = dmabuf->size;
sgl = bfr.sgt->sgl;
buf_addr = (u64)sg_dma_address(bfr.sgt->sgl);
if (g->mm.bypass_smmu || buf_addr == DMA_ERROR_CODE || !buf_addr) {
while (sgl) {
u64 align;
buf_addr = (u64)sg_phys(sgl);
align = 1ULL << __ffs(buf_addr | (u64)sgl->length);
if (bfr.align)
bfr.align = min_t(u64, align, bfr.align);
else
bfr.align = align;
sgl = sg_next(sgl);
}
} else
bfr.align = 1ULL << __ffs(buf_addr);
bfr.pgsz_idx = -1;
mapping_size = mapping_size ? mapping_size : bfr.size;
if (vm->big_pages)
gmmu_select_page_size(vm, &bfr);
else
bfr.pgsz_idx = gmmu_page_size_small;
/* If FIX_OFFSET is set, pgsz is determined at address allocation
* time. The alignment at address alloc time must be the same as
* the alignment determined by gmmu_select_page_size().
*/
if (flags & NVGPU_AS_MAP_BUFFER_FLAGS_FIXED_OFFSET) {
int pgsz_idx =
__nv_gmmu_va_is_big_page_region(vm, offset_align) ?
gmmu_page_size_big : gmmu_page_size_small;
if (pgsz_idx > bfr.pgsz_idx) {
gk20a_err(d, "%llx buffer pgsz %d, VA pgsz %d",
offset_align, bfr.pgsz_idx, pgsz_idx);
err = -EINVAL;
goto clean_up;
}
bfr.pgsz_idx = min(bfr.pgsz_idx, pgsz_idx);
}
/* validate/adjust bfr attributes */
if (unlikely(bfr.pgsz_idx == -1)) {
gk20a_err(d, "unsupported page size detected");
goto clean_up;
}
if (unlikely(bfr.pgsz_idx < gmmu_page_size_small ||
bfr.pgsz_idx > gmmu_page_size_big)) {
BUG_ON(1);
err = -EINVAL;
goto clean_up;
}
gmmu_page_size = vm->gmmu_page_sizes[bfr.pgsz_idx];
/* Check if we should use a fixed offset for mapping this buffer */
if (flags & NVGPU_AS_MAP_BUFFER_FLAGS_FIXED_OFFSET) {
err = validate_fixed_buffer(vm, &bfr,
offset_align, mapping_size,
&va_node);
if (err)
goto clean_up;
map_offset = offset_align;
va_allocated = false;
} else
va_allocated = true;
if (sgt)
*sgt = bfr.sgt;
err = setup_buffer_kind_and_compression(vm, flags, &bfr, bfr.pgsz_idx);
if (unlikely(err)) {
gk20a_err(d, "failure setting up kind and compression");
goto clean_up;
}
/* bar1 and pmu vm don't need ctag */
if (!vm->enable_ctag)
bfr.ctag_lines = 0;
gk20a_get_comptags(d, dmabuf, &comptags);
/* ensure alignment to compression page size if compression enabled */
if (bfr.ctag_offset)
mapping_size = ALIGN(mapping_size,
g->ops.fb.compression_page_size(g));
if (bfr.ctag_lines && !comptags.lines) {
const bool user_mappable =
!!(flags & NVGPU_AS_MAP_BUFFER_FLAGS_MAPPABLE_COMPBITS);
/* allocate compression resources if needed */
err = gk20a_alloc_comptags(g, d, dmabuf, ctag_allocator,
bfr.ctag_lines, user_mappable,
&ctag_map_win_size,
&ctag_map_win_ctagline);
if (err) {
/* ok to fall back here if we ran out */
/* TBD: we can partially alloc ctags as well... */
bfr.kind_v = bfr.uc_kind_v;
} else {
gk20a_get_comptags(d, dmabuf, &comptags);
if (g->ops.ltc.cbc_ctrl)
g->ops.ltc.cbc_ctrl(g, gk20a_cbc_op_clear,
comptags.offset,
comptags.offset +
comptags.allocated_lines - 1);
else
clear_ctags = true;
}
}
/* store the comptag info */
bfr.ctag_offset = comptags.offset;
bfr.ctag_lines = comptags.lines;
bfr.ctag_allocated_lines = comptags.allocated_lines;
bfr.ctag_user_mappable = comptags.user_mappable;
/*
* Calculate comptag index for this mapping. Differs in
* case of partial mapping.
*/
ctag_offset = comptags.offset;
if (ctag_offset)
ctag_offset += buffer_offset >>
ilog2(g->ops.fb.compression_page_size(g));
/* update gmmu ptes */
map_offset = g->ops.mm.gmmu_map(vm, map_offset,
bfr.sgt,
buffer_offset, /* sg offset */
mapping_size,
bfr.pgsz_idx,
bfr.kind_v,
ctag_offset,
flags, rw_flag,
clear_ctags,
false,
false,
batch);
if (!map_offset)
goto clean_up;
gk20a_dbg(gpu_dbg_map,
"as=%d pgsz=%d "
"kind=0x%x kind_uc=0x%x flags=0x%x "
"ctags=%d start=%d ctags_allocated=%d ctags_mappable=%d gv=0x%x,%08x -> 0x%x,%08x -> 0x%x,%08x",
vm_aspace_id(vm), gmmu_page_size,
bfr.kind_v, bfr.uc_kind_v, flags,
bfr.ctag_lines, bfr.ctag_offset,
bfr.ctag_allocated_lines, bfr.ctag_user_mappable,
hi32(map_offset), lo32(map_offset),
hi32((u64)sg_dma_address(bfr.sgt->sgl)),
lo32((u64)sg_dma_address(bfr.sgt->sgl)),
hi32((u64)sg_phys(bfr.sgt->sgl)),
lo32((u64)sg_phys(bfr.sgt->sgl)));
#if defined(NVHOST_DEBUG)
{
int i;
struct scatterlist *sg = NULL;
gk20a_dbg(gpu_dbg_pte, "for_each_sg(bfr.sgt->sgl, sg, bfr.sgt->nents, i)");
for_each_sg(bfr.sgt->sgl, sg, bfr.sgt->nents, i ) {
u64 da = sg_dma_address(sg);
u64 pa = sg_phys(sg);
u64 len = sg->length;
gk20a_dbg(gpu_dbg_pte, "i=%d pa=0x%x,%08x da=0x%x,%08x len=0x%x,%08x",
i, hi32(pa), lo32(pa), hi32(da), lo32(da),
hi32(len), lo32(len));
}
}
#endif
/* keep track of the buffer for unmapping */
/* TBD: check for multiple mapping of same buffer */
mapped_buffer = kzalloc(sizeof(*mapped_buffer), GFP_KERNEL);
if (!mapped_buffer) {
gk20a_warn(d, "oom allocating tracking buffer");
goto clean_up;
}
mapped_buffer->dmabuf = dmabuf;
mapped_buffer->sgt = bfr.sgt;
mapped_buffer->addr = map_offset;
mapped_buffer->size = mapping_size;
mapped_buffer->pgsz_idx = bfr.pgsz_idx;
mapped_buffer->ctag_offset = bfr.ctag_offset;
mapped_buffer->ctag_lines = bfr.ctag_lines;
mapped_buffer->ctag_allocated_lines = bfr.ctag_allocated_lines;
mapped_buffer->ctags_mappable = bfr.ctag_user_mappable;
mapped_buffer->ctag_map_win_size = ctag_map_win_size;
mapped_buffer->ctag_map_win_ctagline = ctag_map_win_ctagline;
mapped_buffer->vm = vm;
mapped_buffer->flags = flags;
mapped_buffer->kind = kind;
mapped_buffer->va_allocated = va_allocated;
mapped_buffer->user_mapped = user_mapped ? 1 : 0;
mapped_buffer->own_mem_ref = user_mapped;
INIT_LIST_HEAD(&mapped_buffer->unmap_list);
INIT_LIST_HEAD(&mapped_buffer->va_buffers_list);
kref_init(&mapped_buffer->ref);
err = insert_mapped_buffer(&vm->mapped_buffers, mapped_buffer);
if (err) {
gk20a_err(d, "failed to insert into mapped buffer tree");
goto clean_up;
}
inserted = true;
if (user_mapped)
vm->num_user_mapped_buffers++;
gk20a_dbg_info("allocated va @ 0x%llx", map_offset);
if (va_node) {
list_add_tail(&mapped_buffer->va_buffers_list,
&va_node->va_buffers_list);
mapped_buffer->va_node = va_node;
}
mutex_unlock(&vm->update_gmmu_lock);
return map_offset;
clean_up:
if (inserted) {
rb_erase(&mapped_buffer->node, &vm->mapped_buffers);
if (user_mapped)
vm->num_user_mapped_buffers--;
}
kfree(mapped_buffer);
if (va_allocated)
gk20a_vm_free_va(vm, map_offset, bfr.size, bfr.pgsz_idx);
if (!IS_ERR(bfr.sgt))
gk20a_mm_unpin(d, dmabuf, bfr.sgt);
mutex_unlock(&vm->update_gmmu_lock);
gk20a_dbg_info("err=%d\n", err);
return 0;
}
int gk20a_vm_get_compbits_info(struct vm_gk20a *vm,
u64 mapping_gva,
u64 *compbits_win_size,
u32 *compbits_win_ctagline,
u32 *mapping_ctagline,
u32 *flags)
{
struct mapped_buffer_node *mapped_buffer;
struct device *d = dev_from_vm(vm);
mutex_lock(&vm->update_gmmu_lock);
mapped_buffer = find_mapped_buffer_locked(&vm->mapped_buffers, mapping_gva);
if (!mapped_buffer || !mapped_buffer->user_mapped)
{
mutex_unlock(&vm->update_gmmu_lock);
gk20a_err(d, "%s: bad offset 0x%llx", __func__, mapping_gva);
return -EFAULT;
}
*compbits_win_size = 0;
*compbits_win_ctagline = 0;
*mapping_ctagline = 0;
*flags = 0;
if (mapped_buffer->ctag_offset)
*flags |= NVGPU_AS_GET_BUFFER_COMPBITS_INFO_FLAGS_HAS_COMPBITS;
if (mapped_buffer->ctags_mappable)
{
*flags |= NVGPU_AS_GET_BUFFER_COMPBITS_INFO_FLAGS_MAPPABLE;
*compbits_win_size = mapped_buffer->ctag_map_win_size;
*compbits_win_ctagline = mapped_buffer->ctag_map_win_ctagline;
*mapping_ctagline = mapped_buffer->ctag_offset;
}
mutex_unlock(&vm->update_gmmu_lock);
return 0;
}
int gk20a_vm_map_compbits(struct vm_gk20a *vm,
u64 mapping_gva,
u64 *compbits_win_gva,
u64 *mapping_iova,
u32 flags)
{
struct mapped_buffer_node *mapped_buffer;
struct gk20a *g = gk20a_from_vm(vm);
struct device *d = dev_from_vm(vm);
const bool fixed_mapping =
(flags & NVGPU_AS_MAP_BUFFER_COMPBITS_FLAGS_FIXED_OFFSET) != 0;
if (vm->userspace_managed && !fixed_mapping) {
gk20a_err(d,
"%s: non-fixed-offset mapping is not available on userspace managed address spaces",
__func__);
return -EFAULT;
}
if (fixed_mapping && !vm->userspace_managed) {
gk20a_err(d,
"%s: fixed-offset mapping is available only on userspace managed address spaces",
__func__);
return -EFAULT;
}
mutex_lock(&vm->update_gmmu_lock);
mapped_buffer =
find_mapped_buffer_locked(&vm->mapped_buffers, mapping_gva);
if (!mapped_buffer || !mapped_buffer->user_mapped) {
mutex_unlock(&vm->update_gmmu_lock);
gk20a_err(d, "%s: bad offset 0x%llx", __func__, mapping_gva);
return -EFAULT;
}
if (!mapped_buffer->ctags_mappable) {
mutex_unlock(&vm->update_gmmu_lock);
gk20a_err(d, "%s: comptags not mappable, offset 0x%llx",
__func__, mapping_gva);
return -EFAULT;
}
if (!mapped_buffer->ctag_map_win_addr) {
const u32 small_pgsz_index = 0; /* small pages, 4K */
const u32 aggregate_cacheline_sz =
g->gr.cacheline_size * g->gr.slices_per_ltc *
g->ltc_count;
/* first aggregate cacheline to map */
u32 cacheline_start; /* inclusive */
/* offset of the start cacheline (will be page aligned) */
u64 cacheline_offset_start;
if (!mapped_buffer->ctag_map_win_size) {
mutex_unlock(&vm->update_gmmu_lock);
gk20a_err(d,
"%s: mapping 0x%llx does not have "
"mappable comptags",
__func__, mapping_gva);
return -EFAULT;
}
cacheline_start = mapped_buffer->ctag_offset /
g->gr.comptags_per_cacheline;
cacheline_offset_start =
cacheline_start * aggregate_cacheline_sz;
if (fixed_mapping) {
struct buffer_attrs bfr;
int err;
struct vm_reserved_va_node *va_node = NULL;
memset(&bfr, 0, sizeof(bfr));
bfr.pgsz_idx = small_pgsz_index;
err = validate_fixed_buffer(
vm, &bfr, *compbits_win_gva,
mapped_buffer->ctag_map_win_size, &va_node);
if (err) {
mutex_unlock(&vm->update_gmmu_lock);
return err;
}
if (va_node) {
/* this would create a dangling GPU VA
* pointer if the space is freed
* before before the buffer is
* unmapped */
mutex_unlock(&vm->update_gmmu_lock);
gk20a_err(d,
"%s: comptags cannot be mapped into allocated space",
__func__);
return -EINVAL;
}
}
mapped_buffer->ctag_map_win_addr =
g->ops.mm.gmmu_map(
vm,
!fixed_mapping ? 0 : *compbits_win_gva, /* va */
g->gr.compbit_store.mem.sgt,
cacheline_offset_start, /* sg offset */
mapped_buffer->ctag_map_win_size, /* size */
small_pgsz_index,
0, /* kind */
0, /* ctag_offset */
NVGPU_MAP_BUFFER_FLAGS_CACHEABLE_TRUE,
gk20a_mem_flag_read_only,
false, /* clear_ctags */
false, /* sparse */
false, /* priv */
NULL); /* mapping_batch handle */
if (!mapped_buffer->ctag_map_win_addr) {
mutex_unlock(&vm->update_gmmu_lock);
gk20a_err(d,
"%s: failed to map comptags for mapping 0x%llx",
__func__, mapping_gva);
return -ENOMEM;
}
} else if (fixed_mapping && *compbits_win_gva &&
mapped_buffer->ctag_map_win_addr != *compbits_win_gva) {
mutex_unlock(&vm->update_gmmu_lock);
gk20a_err(d,
"%s: re-requesting comptags map into mismatching address. buffer offset 0x"
"%llx, existing comptag map at 0x%llx, requested remap 0x%llx",
__func__, mapping_gva,
mapped_buffer->ctag_map_win_addr, *compbits_win_gva);
return -EINVAL;
}
*mapping_iova = gk20a_mm_iova_addr(g, mapped_buffer->sgt->sgl, 0);
*compbits_win_gva = mapped_buffer->ctag_map_win_addr;
mutex_unlock(&vm->update_gmmu_lock);
return 0;
}
/*
* Core GMMU map function for the kernel to use. If @addr is 0 then the GPU
* VA will be allocated for you. If addr is non-zero then the buffer will be
* mapped at @addr.
*/
static u64 __gk20a_gmmu_map(struct vm_gk20a *vm,
struct sg_table **sgt,
u64 addr,
u64 size,
u32 flags,
int rw_flag,
bool priv)
{
struct gk20a *g = gk20a_from_vm(vm);
u64 vaddr;
mutex_lock(&vm->update_gmmu_lock);
vaddr = g->ops.mm.gmmu_map(vm, addr,
*sgt, /* sg table */
0, /* sg offset */
size,
gmmu_page_size_kernel,
0, /* kind */
0, /* ctag_offset */
flags, rw_flag,
false, /* clear_ctags */
false, /* sparse */
priv, /* priv */
NULL); /* mapping_batch handle */
mutex_unlock(&vm->update_gmmu_lock);
if (!vaddr) {
gk20a_err(dev_from_vm(vm), "failed to allocate va space");
return 0;
}
return vaddr;
}
u64 gk20a_gmmu_map(struct vm_gk20a *vm,
struct sg_table **sgt,
u64 size,
u32 flags,
int rw_flag,
bool priv)
{
return __gk20a_gmmu_map(vm, sgt, 0, size, flags, rw_flag, priv);
}
/*
* Like gk20a_gmmu_map() except it works on a fixed address instead.
*/
u64 gk20a_gmmu_fixed_map(struct vm_gk20a *vm,
struct sg_table **sgt,
u64 addr,
u64 size,
u32 flags,
int rw_flag,
bool priv)
{
return __gk20a_gmmu_map(vm, sgt, addr, size, flags, rw_flag, priv);
}
int gk20a_gmmu_alloc(struct gk20a *g, size_t size, struct mem_desc *mem)
{
return gk20a_gmmu_alloc_attr(g, 0, size, mem);
}
int gk20a_gmmu_alloc_attr(struct gk20a *g, enum dma_attr attr, size_t size, struct mem_desc *mem)
{
struct device *d = dev_from_gk20a(g);
int err;
dma_addr_t iova;
gk20a_dbg_fn("");
if (attr) {
DEFINE_DMA_ATTRS(attrs);
dma_set_attr(attr, &attrs);
if (attr == DMA_ATTR_NO_KERNEL_MAPPING) {
mem->pages = dma_alloc_attrs(d,
size, &iova, GFP_KERNEL, &attrs);
if (!mem->pages)
return -ENOMEM;
} else {
mem->cpu_va = dma_alloc_attrs(d,
size, &iova, GFP_KERNEL, &attrs);
if (!mem->cpu_va)
return -ENOMEM;
}
} else {
mem->cpu_va = dma_alloc_coherent(d, size, &iova, GFP_KERNEL);
if (!mem->cpu_va)
return -ENOMEM;
}
if (attr == DMA_ATTR_NO_KERNEL_MAPPING)
err = gk20a_get_sgtable_from_pages(d, &mem->sgt, mem->pages,
iova, size);
else {
err = gk20a_get_sgtable(d, &mem->sgt, mem->cpu_va, iova, size);
memset(mem->cpu_va, 0, size);
}
if (err)
goto fail_free;
mem->size = size;
gk20a_dbg_fn("done");
return 0;
fail_free:
dma_free_coherent(d, size, mem->cpu_va, iova);
mem->cpu_va = NULL;
mem->sgt = NULL;
return err;
}
void gk20a_gmmu_free_attr(struct gk20a *g, enum dma_attr attr,
struct mem_desc *mem)
{
struct device *d = dev_from_gk20a(g);
if (mem->cpu_va || mem->pages) {
if (attr) {
DEFINE_DMA_ATTRS(attrs);
dma_set_attr(attr, &attrs);
if (attr == DMA_ATTR_NO_KERNEL_MAPPING) {
if (mem->pages)
dma_free_attrs(d, mem->size, mem->pages,
sg_dma_address(mem->sgt->sgl),
&attrs);
} else {
if (mem->cpu_va)
dma_free_attrs(d, mem->size,
mem->cpu_va,
sg_dma_address(mem->sgt->sgl),
&attrs);
}
} else {
if (mem->cpu_va)
dma_free_coherent(d, mem->size, mem->cpu_va,
sg_dma_address(mem->sgt->sgl));
}
mem->cpu_va = NULL;
mem->pages = NULL;
}
if (mem->sgt)
gk20a_free_sgtable(&mem->sgt);
mem->size = 0;
}
void gk20a_gmmu_free(struct gk20a *g, struct mem_desc *mem)
{
return gk20a_gmmu_free_attr(g, 0, mem);
}
int gk20a_gmmu_alloc_map(struct vm_gk20a *vm, size_t size, struct mem_desc *mem)
{
return gk20a_gmmu_alloc_map_attr(vm, 0, size, mem);
}
int gk20a_gmmu_alloc_map_attr(struct vm_gk20a *vm,
enum dma_attr attr, size_t size, struct mem_desc *mem)
{
int err = gk20a_gmmu_alloc_attr(vm->mm->g, attr, size, mem);
if (err)
return err;
mem->gpu_va = gk20a_gmmu_map(vm, &mem->sgt, size, 0,
gk20a_mem_flag_none, false);
if (!mem->gpu_va) {
err = -ENOMEM;
goto fail_free;
}
return 0;
fail_free:
gk20a_gmmu_free(vm->mm->g, mem);
return err;
}
void gk20a_gmmu_unmap_free(struct vm_gk20a *vm, struct mem_desc *mem)
{
if (mem->gpu_va)
gk20a_gmmu_unmap(vm, mem->gpu_va, mem->size, gk20a_mem_flag_none);
mem->gpu_va = 0;
gk20a_gmmu_free(vm->mm->g, mem);
}
dma_addr_t gk20a_mm_gpuva_to_iova_base(struct vm_gk20a *vm, u64 gpu_vaddr)
{
struct mapped_buffer_node *buffer;
dma_addr_t addr = 0;
struct gk20a *g = gk20a_from_vm(vm);
mutex_lock(&vm->update_gmmu_lock);
buffer = find_mapped_buffer_locked(&vm->mapped_buffers, gpu_vaddr);
if (buffer)
addr = g->ops.mm.get_iova_addr(g, buffer->sgt->sgl,
buffer->flags);
mutex_unlock(&vm->update_gmmu_lock);
return addr;
}
void gk20a_gmmu_unmap(struct vm_gk20a *vm,
u64 vaddr,
u64 size,
int rw_flag)
{
struct gk20a *g = gk20a_from_vm(vm);
mutex_lock(&vm->update_gmmu_lock);
g->ops.mm.gmmu_unmap(vm,
vaddr,
size,
gmmu_page_size_kernel,
true, /*va_allocated */
rw_flag,
false,
NULL);
mutex_unlock(&vm->update_gmmu_lock);
}
phys_addr_t gk20a_get_phys_from_iova(struct device *d,
u64 dma_addr)
{
phys_addr_t phys;
u64 iova;
struct dma_iommu_mapping *mapping = to_dma_iommu_mapping(d);
if (!mapping)
return dma_addr;
iova = dma_addr & PAGE_MASK;
phys = iommu_iova_to_phys(mapping->domain, iova);
return phys;
}
/* get sg_table from already allocated buffer */
int gk20a_get_sgtable(struct device *d, struct sg_table **sgt,
void *cpuva, u64 iova,
size_t size)
{
int err = 0;
*sgt = kzalloc(sizeof(struct sg_table), GFP_KERNEL);
if (!(*sgt)) {
dev_err(d, "failed to allocate memory\n");
err = -ENOMEM;
goto fail;
}
err = dma_get_sgtable(d, *sgt,
cpuva, iova,
size);
if (err) {
dev_err(d, "failed to create sg table\n");
goto fail;
}
sg_dma_address((*sgt)->sgl) = iova;
return 0;
fail:
if (*sgt) {
kfree(*sgt);
*sgt = NULL;
}
return err;
}
int gk20a_get_sgtable_from_pages(struct device *d, struct sg_table **sgt,
struct page **pages, u64 iova,
size_t size)
{
int err = 0;
*sgt = kzalloc(sizeof(struct sg_table), GFP_KERNEL);
if (!(*sgt)) {
dev_err(d, "failed to allocate memory\n");
err = -ENOMEM;
goto fail;
}
err = sg_alloc_table_from_pages(*sgt, pages,
DIV_ROUND_UP(size, 4096), 0, size, GFP_KERNEL);
if (err) {
dev_err(d, "failed to allocate sg_table\n");
goto fail;
}
sg_dma_address((*sgt)->sgl) = iova;
return 0;
fail:
if (*sgt) {
kfree(*sgt);
*sgt = NULL;
}
return err;
}
void gk20a_free_sgtable(struct sg_table **sgt)
{
sg_free_table(*sgt);
kfree(*sgt);
*sgt = NULL;
}
u64 gk20a_mm_smmu_vaddr_translate(struct gk20a *g, dma_addr_t iova)
{
if (!device_is_iommuable(dev_from_gk20a(g)) || !g->mm.has_physical_mode)
return iova;
else
return iova | 1ULL << g->ops.mm.get_physical_addr_bits(g);
}
u64 gk20a_mm_iova_addr(struct gk20a *g, struct scatterlist *sgl,
u32 flags)
{
if (!device_is_iommuable(dev_from_gk20a(g)))
return sg_phys(sgl);
if (sg_dma_address(sgl) == 0)
return sg_phys(sgl);
if (sg_dma_address(sgl) == DMA_ERROR_CODE)
return 0;
return gk20a_mm_smmu_vaddr_translate(g, sg_dma_address(sgl));
}
/* for gk20a the "video memory" apertures here are misnomers. */
static inline u32 big_valid_pde0_bits(struct gk20a *g, u64 pte_addr)
{
u32 pde0_bits =
(g->mm.vidmem_is_vidmem ?
gmmu_pde_aperture_big_sys_mem_ncoh_f() :
gmmu_pde_aperture_big_video_memory_f()) |
gmmu_pde_address_big_sys_f(
(u32)(pte_addr >> gmmu_pde_address_shift_v()));
return pde0_bits;
}
static inline u32 small_valid_pde1_bits(struct gk20a *g, u64 pte_addr)
{
u32 pde1_bits =
(g->mm.vidmem_is_vidmem ?
gmmu_pde_aperture_small_sys_mem_ncoh_f() :
gmmu_pde_aperture_small_video_memory_f()) |
gmmu_pde_vol_small_true_f() | /* tbd: why? */
gmmu_pde_address_small_sys_f(
(u32)(pte_addr >> gmmu_pde_address_shift_v()));
return pde1_bits;
}
/* Given the current state of the ptes associated with a pde,
determine value and write it out. There's no checking
here to determine whether or not a change was actually
made. So, superfluous updates will cause unnecessary
pde invalidations.
*/
static int update_gmmu_pde_locked(struct vm_gk20a *vm,
struct gk20a_mm_entry *pte,
u32 i, u32 gmmu_pgsz_idx,
struct scatterlist **sgl,
u64 *offset,
u64 *iova,
u32 kind_v, u64 *ctag,
bool cacheable, bool unammped_pte,
int rw_flag, bool sparse, bool priv)
{
struct gk20a *g = gk20a_from_vm(vm);
bool small_valid, big_valid;
u64 pte_addr_small = 0, pte_addr_big = 0;
struct gk20a_mm_entry *entry = vm->pdb.entries + i;
u32 pde_v[2] = {0, 0};
u32 pde;
gk20a_dbg_fn("");
small_valid = entry->mem.size && entry->pgsz == gmmu_page_size_small;
big_valid = entry->mem.size && entry->pgsz == gmmu_page_size_big;
if (small_valid)
pte_addr_small = g->ops.mm.get_iova_addr(g, entry->mem.sgt->sgl, 0);
if (big_valid)
pte_addr_big = g->ops.mm.get_iova_addr(g, entry->mem.sgt->sgl, 0);
pde_v[0] = gmmu_pde_size_full_f();
pde_v[0] |= big_valid ? big_valid_pde0_bits(g, pte_addr_big) :
(gmmu_pde_aperture_big_invalid_f());
pde_v[1] |= (small_valid ?
small_valid_pde1_bits(g, pte_addr_small) :
(gmmu_pde_aperture_small_invalid_f() |
gmmu_pde_vol_small_false_f()))
|
(big_valid ? (gmmu_pde_vol_big_true_f()) :
gmmu_pde_vol_big_false_f());
pde = pde_from_index(i);
gk20a_mem_wr32(g, &vm->pdb.mem, pde + 0, pde_v[0]);
gk20a_mem_wr32(g, &vm->pdb.mem, pde + 1, pde_v[1]);
gk20a_dbg(gpu_dbg_pte, "pde:%d,sz=%d = 0x%x,0x%08x",
i, gmmu_pgsz_idx, pde_v[1], pde_v[0]);
return 0;
}
static int update_gmmu_pte_locked(struct vm_gk20a *vm,
struct gk20a_mm_entry *pte,
u32 i, u32 gmmu_pgsz_idx,
struct scatterlist **sgl,
u64 *offset,
u64 *iova,
u32 kind_v, u64 *ctag,
bool cacheable, bool unmapped_pte,
int rw_flag, bool sparse, bool priv)
{
struct gk20a *g = gk20a_from_vm(vm);
int ctag_shift = ilog2(g->ops.fb.compression_page_size(g));
u32 page_size = vm->gmmu_page_sizes[gmmu_pgsz_idx];
u32 pte_w[2] = {0, 0}; /* invalid pte */
if (*iova) {
if (unmapped_pte)
pte_w[0] = gmmu_pte_valid_false_f() |
gmmu_pte_address_sys_f(*iova
>> gmmu_pte_address_shift_v());
else
pte_w[0] = gmmu_pte_valid_true_f() |
gmmu_pte_address_sys_f(*iova
>> gmmu_pte_address_shift_v());
if (priv)
pte_w[0] |= gmmu_pte_privilege_true_f();
pte_w[1] = (g->mm.vidmem_is_vidmem ?
gmmu_pte_aperture_sys_mem_ncoh_f() :
gmmu_pte_aperture_video_memory_f()) |
gmmu_pte_kind_f(kind_v) |
gmmu_pte_comptagline_f((u32)(*ctag >> ctag_shift));
if (*ctag && vm->mm->use_full_comp_tag_line && *iova & 0x10000)
pte_w[1] |= gmmu_pte_comptagline_f(
1 << (gmmu_pte_comptagline_s() - 1));
if (rw_flag == gk20a_mem_flag_read_only) {
pte_w[0] |= gmmu_pte_read_only_true_f();
pte_w[1] |=
gmmu_pte_write_disable_true_f();
} else if (rw_flag ==
gk20a_mem_flag_write_only) {
pte_w[1] |=
gmmu_pte_read_disable_true_f();
}
if (!unmapped_pte) {
if (!cacheable)
pte_w[1] |=
gmmu_pte_vol_true_f();
} else {
/* Store cacheable value behind
* gmmu_pte_write_disable_true_f */
if (!cacheable)
pte_w[1] |=
gmmu_pte_write_disable_true_f();
}
gk20a_dbg(gpu_dbg_pte,
"pte=%d iova=0x%llx kind=%d ctag=%d vol=%d [0x%08x, 0x%08x]",
i, *iova,
kind_v, (u32)(*ctag >> ctag_shift), !cacheable,
pte_w[1], pte_w[0]);
if (*ctag)
*ctag += page_size;
} else if (sparse) {
pte_w[0] = gmmu_pte_valid_false_f();
pte_w[1] |= gmmu_pte_vol_true_f();
} else {
gk20a_dbg(gpu_dbg_pte, "pte_cur=%d [0x0,0x0]", i);
}
gk20a_mem_wr32(g, &pte->mem, pte_from_index(i) + 0, pte_w[0]);
gk20a_mem_wr32(g, &pte->mem, pte_from_index(i) + 1, pte_w[1]);
if (*iova) {
*iova += page_size;
*offset += page_size;
if (*sgl && *offset + page_size > (*sgl)->length) {
u64 new_iova;
*sgl = sg_next(*sgl);
if (*sgl) {
new_iova = sg_phys(*sgl);
gk20a_dbg(gpu_dbg_pte, "chunk address %llx, size %d",
new_iova, (*sgl)->length);
if (new_iova) {
*offset = 0;
*iova = new_iova;
}
}
}
}
return 0;
}
static int update_gmmu_level_locked(struct vm_gk20a *vm,
struct gk20a_mm_entry *pte,
enum gmmu_pgsz_gk20a pgsz_idx,
struct scatterlist **sgl,
u64 *offset,
u64 *iova,
u64 gpu_va, u64 gpu_end,
u8 kind_v, u64 *ctag,
bool cacheable, bool unmapped_pte,
int rw_flag,
bool sparse,
int lvl,
bool priv)
{
const struct gk20a_mmu_level *l = &vm->mmu_levels[lvl];
const struct gk20a_mmu_level *next_l = &vm->mmu_levels[lvl+1];
int err = 0;
u32 pde_i;
u64 pde_size = 1ULL << (u64)l->lo_bit[pgsz_idx];
gk20a_dbg_fn("");
pde_i = (gpu_va & ((1ULL << ((u64)l->hi_bit[pgsz_idx]+1)) - 1ULL))
>> (u64)l->lo_bit[pgsz_idx];
gk20a_dbg(gpu_dbg_pte, "size_idx=%d, l: %d, [%llx,%llx], iova=%llx",
pgsz_idx, lvl, gpu_va, gpu_end-1, *iova);
while (gpu_va < gpu_end) {
struct gk20a_mm_entry *next_pte = NULL;
u64 next = min((gpu_va + pde_size) & ~(pde_size-1), gpu_end);
/* Allocate next level */
if (next_l->update_entry) {
if (!pte->entries) {
int num_entries =
1 <<
(l->hi_bit[pgsz_idx]
- l->lo_bit[pgsz_idx] + 1);
pte->entries =
vzalloc(sizeof(struct gk20a_mm_entry) *
num_entries);
if (!pte->entries)
return -ENOMEM;
pte->pgsz = pgsz_idx;
pte->num_entries = num_entries;
}
next_pte = pte->entries + pde_i;
if (!next_pte->mem.size) {
err = gk20a_zalloc_gmmu_page_table(vm,
pgsz_idx, next_l, next_pte);
if (err)
return err;
}
}
err = l->update_entry(vm, pte, pde_i, pgsz_idx,
sgl, offset, iova,
kind_v, ctag, cacheable, unmapped_pte,
rw_flag, sparse, priv);
if (err)
return err;
if (next_l->update_entry) {
/* get cpu access to the ptes */
err = map_gmmu_pages(next_pte);
if (err) {
gk20a_err(dev_from_vm(vm),
"couldn't map ptes for update as=%d",
vm_aspace_id(vm));
return err;
}
err = update_gmmu_level_locked(vm, next_pte,
pgsz_idx,
sgl,
offset,
iova,
gpu_va,
next,
kind_v, ctag, cacheable, unmapped_pte,
rw_flag, sparse, lvl+1, priv);
unmap_gmmu_pages(next_pte);
if (err)
return err;
}
pde_i++;
gpu_va = next;
}
gk20a_dbg_fn("done");
return 0;
}
static int update_gmmu_ptes_locked(struct vm_gk20a *vm,
enum gmmu_pgsz_gk20a pgsz_idx,
struct sg_table *sgt,
u64 buffer_offset,
u64 gpu_va, u64 gpu_end,
u8 kind_v, u32 ctag_offset,
bool cacheable, bool unmapped_pte,
int rw_flag,
bool sparse,
bool priv)
{
struct gk20a *g = gk20a_from_vm(vm);
int ctag_granularity = g->ops.fb.compression_page_size(g);
u64 ctag = (u64)ctag_offset * (u64)ctag_granularity;
u64 iova = 0;
u64 space_to_skip = buffer_offset;
u32 page_size = vm->gmmu_page_sizes[pgsz_idx];
int err;
struct scatterlist *sgl = NULL;
gk20a_dbg(gpu_dbg_pte, "size_idx=%d, iova=%llx, buffer offset %lld, nents %d",
pgsz_idx,
sgt ? g->ops.mm.get_iova_addr(vm->mm->g, sgt->sgl, 0)
: 0ULL,
buffer_offset,
sgt ? sgt->nents : 0);
/* note: here we need to map kernel to small, since the
* low-level mmu code assumes 0 is small and 1 is big pages */
if (pgsz_idx == gmmu_page_size_kernel)
pgsz_idx = gmmu_page_size_small;
if (space_to_skip & (page_size - 1))
return -EINVAL;
if (sgt) {
iova = g->ops.mm.get_iova_addr(vm->mm->g, sgt->sgl, 0);
if (!vm->mm->bypass_smmu && iova) {
iova += space_to_skip;
} else {
sgl = sgt->sgl;
gk20a_dbg(gpu_dbg_pte, "chunk address %llx, size %d",
(u64)sg_phys(sgl),
sgl->length);
while (space_to_skip && sgl &&
space_to_skip + page_size > sgl->length) {
space_to_skip -= sgl->length;
sgl = sg_next(sgl);
gk20a_dbg(gpu_dbg_pte, "chunk address %llx, size %d",
(u64)sg_phys(sgl),
sgl->length);
}
iova = sg_phys(sgl) + space_to_skip;
}
}
gk20a_dbg(gpu_dbg_map, "size_idx=%d, gpu_va=[%llx,%llx], iova=%llx",
pgsz_idx, gpu_va, gpu_end-1, iova);
err = map_gmmu_pages(&vm->pdb);
if (err) {
gk20a_err(dev_from_vm(vm),
"couldn't map ptes for update as=%d",
vm_aspace_id(vm));
return err;
}
err = update_gmmu_level_locked(vm, &vm->pdb, pgsz_idx,
&sgl,
&space_to_skip,
&iova,
gpu_va, gpu_end,
kind_v, &ctag,
cacheable, unmapped_pte, rw_flag, sparse, 0, priv);
unmap_gmmu_pages(&vm->pdb);
smp_mb();
gk20a_dbg_fn("done");
return err;
}
/* NOTE! mapped_buffers lock must be held */
void gk20a_vm_unmap_locked(struct mapped_buffer_node *mapped_buffer,
struct vm_gk20a_mapping_batch *batch)
{
struct vm_gk20a *vm = mapped_buffer->vm;
struct gk20a *g = vm->mm->g;
if (mapped_buffer->ctag_map_win_addr) {
/* unmap compbits */
g->ops.mm.gmmu_unmap(vm,
mapped_buffer->ctag_map_win_addr,
mapped_buffer->ctag_map_win_size,
0, /* page size 4k */
true, /* va allocated */
gk20a_mem_flag_none,
false, /* not sparse */
batch); /* batch handle */
}
g->ops.mm.gmmu_unmap(vm,
mapped_buffer->addr,
mapped_buffer->size,
mapped_buffer->pgsz_idx,
mapped_buffer->va_allocated,
gk20a_mem_flag_none,
mapped_buffer->va_node ?
mapped_buffer->va_node->sparse : false,
batch);
gk20a_dbg(gpu_dbg_map, "as=%d pgsz=%d gv=0x%x,%08x own_mem_ref=%d",
vm_aspace_id(vm),
vm->gmmu_page_sizes[mapped_buffer->pgsz_idx],
hi32(mapped_buffer->addr), lo32(mapped_buffer->addr),
mapped_buffer->own_mem_ref);
gk20a_mm_unpin(dev_from_vm(vm), mapped_buffer->dmabuf,
mapped_buffer->sgt);
/* remove from mapped buffer tree and remove list, free */
rb_erase(&mapped_buffer->node, &vm->mapped_buffers);
if (!list_empty(&mapped_buffer->va_buffers_list))
list_del(&mapped_buffer->va_buffers_list);
/* keep track of mapped buffers */
if (mapped_buffer->user_mapped)
vm->num_user_mapped_buffers--;
if (mapped_buffer->own_mem_ref)
dma_buf_put(mapped_buffer->dmabuf);
kfree(mapped_buffer);
return;
}
void gk20a_vm_unmap(struct vm_gk20a *vm, u64 offset)
{
struct device *d = dev_from_vm(vm);
struct mapped_buffer_node *mapped_buffer;
mutex_lock(&vm->update_gmmu_lock);
mapped_buffer = find_mapped_buffer_locked(&vm->mapped_buffers, offset);
if (!mapped_buffer) {
mutex_unlock(&vm->update_gmmu_lock);
gk20a_err(d, "invalid addr to unmap 0x%llx", offset);
return;
}
kref_put(&mapped_buffer->ref, gk20a_vm_unmap_locked_kref);
mutex_unlock(&vm->update_gmmu_lock);
}
static void gk20a_vm_free_entries(struct vm_gk20a *vm,
struct gk20a_mm_entry *parent,
int level)
{
int i;
if (parent->entries)
for (i = 0; i < parent->num_entries; i++)
gk20a_vm_free_entries(vm, &parent->entries[i], level+1);
if (parent->mem.size)
free_gmmu_pages(vm, parent);
vfree(parent->entries);
parent->entries = NULL;
}
static void gk20a_vm_remove_support_nofree(struct vm_gk20a *vm)
{
struct mapped_buffer_node *mapped_buffer;
struct vm_reserved_va_node *va_node, *va_node_tmp;
struct rb_node *node;
gk20a_dbg_fn("");
mutex_lock(&vm->update_gmmu_lock);
/* TBD: add a flag here for the unmap code to recognize teardown
* and short-circuit any otherwise expensive operations. */
node = rb_first(&vm->mapped_buffers);
while (node) {
mapped_buffer =
container_of(node, struct mapped_buffer_node, node);
gk20a_vm_unmap_locked(mapped_buffer, NULL);
node = rb_first(&vm->mapped_buffers);
}
/* destroy remaining reserved memory areas */
list_for_each_entry_safe(va_node, va_node_tmp, &vm->reserved_va_list,
reserved_va_list) {
list_del(&va_node->reserved_va_list);
kfree(va_node);
}
gk20a_deinit_vm(vm);
mutex_unlock(&vm->update_gmmu_lock);
}
void gk20a_vm_remove_support(struct vm_gk20a *vm)
{
gk20a_vm_remove_support_nofree(vm);
/* vm is not used anymore. release it. */
kfree(vm);
}
static void gk20a_vm_remove_support_kref(struct kref *ref)
{
struct vm_gk20a *vm = container_of(ref, struct vm_gk20a, ref);
struct gk20a *g = gk20a_from_vm(vm);
g->ops.mm.vm_remove(vm);
}
void gk20a_vm_get(struct vm_gk20a *vm)
{
kref_get(&vm->ref);
}
void gk20a_vm_put(struct vm_gk20a *vm)
{
kref_put(&vm->ref, gk20a_vm_remove_support_kref);
}
const struct gk20a_mmu_level gk20a_mm_levels_64k[] = {
{.hi_bit = {NV_GMMU_VA_RANGE-1, NV_GMMU_VA_RANGE-1},
.lo_bit = {26, 26},
.update_entry = update_gmmu_pde_locked,
.entry_size = 8},
{.hi_bit = {25, 25},
.lo_bit = {12, 16},
.update_entry = update_gmmu_pte_locked,
.entry_size = 8},
{.update_entry = NULL}
};
const struct gk20a_mmu_level gk20a_mm_levels_128k[] = {
{.hi_bit = {NV_GMMU_VA_RANGE-1, NV_GMMU_VA_RANGE-1},
.lo_bit = {27, 27},
.update_entry = update_gmmu_pde_locked,
.entry_size = 8},
{.hi_bit = {26, 26},
.lo_bit = {12, 17},
.update_entry = update_gmmu_pte_locked,
.entry_size = 8},
{.update_entry = NULL}
};
int gk20a_init_vm(struct mm_gk20a *mm,
struct vm_gk20a *vm,
u32 big_page_size,
u64 low_hole,
u64 kernel_reserved,
u64 aperture_size,
bool big_pages,
bool userspace_managed,
char *name)
{
int err, i;
char alloc_name[32];
u64 small_vma_start, small_vma_limit, large_vma_start, large_vma_limit,
kernel_vma_start, kernel_vma_limit;
u32 pde_lo, pde_hi;
struct gk20a *g = mm->g;
/* note: this must match gmmu_pgsz_gk20a enum */
u32 gmmu_page_sizes[gmmu_nr_page_sizes] = { SZ_4K, big_page_size, SZ_4K };
WARN_ON(kernel_reserved + low_hole > aperture_size);
if (kernel_reserved > aperture_size)
return -ENOMEM;
vm->mm = mm;
vm->va_start = low_hole;
vm->va_limit = aperture_size;
vm->big_pages = big_pages;
vm->big_page_size = gmmu_page_sizes[gmmu_page_size_big];
vm->userspace_managed = userspace_managed;
vm->mmu_levels = vm->mm->g->ops.mm.get_mmu_levels(vm->mm->g,
vm->big_page_size);
for (i = 0; i < gmmu_nr_page_sizes; i++)
vm->gmmu_page_sizes[i] = gmmu_page_sizes[i];
gk20a_dbg_info("small page-size (%dKB)",
vm->gmmu_page_sizes[gmmu_page_size_small] >> 10);
gk20a_dbg_info("big page-size (%dKB)",
vm->gmmu_page_sizes[gmmu_page_size_big] >> 10);
gk20a_dbg_info("kernel page-size (%dKB)",
vm->gmmu_page_sizes[gmmu_page_size_kernel] >> 10);
pde_range_from_vaddr_range(vm,
0, vm->va_limit-1,
&pde_lo, &pde_hi);
vm->pdb.entries = vzalloc(sizeof(struct gk20a_mm_entry) *
(pde_hi + 1));
vm->pdb.num_entries = pde_hi + 1;
if (!vm->pdb.entries)
return -ENOMEM;
gk20a_dbg_info("init space for %s va_limit=0x%llx num_pdes=%d",
name, vm->va_limit, pde_hi + 1);
/* allocate the page table directory */
err = gk20a_zalloc_gmmu_page_table(vm, 0, &vm->mmu_levels[0], &vm->pdb);
if (err)
goto clean_up_pdes;
/* setup vma limits */
small_vma_start = low_hole;
if (big_pages) {
/* First 16GB of the address space goes towards small
* pages. What ever remains is allocated to large
* pages. */
small_vma_limit = __nv_gmmu_va_small_page_limit();
large_vma_start = small_vma_limit;
large_vma_limit = vm->va_limit - kernel_reserved;
} else {
small_vma_limit = vm->va_limit - kernel_reserved;
large_vma_start = 0;
large_vma_limit = 0;
}
kernel_vma_start = vm->va_limit - kernel_reserved;
kernel_vma_limit = vm->va_limit;
gk20a_dbg_info(
"small_vma=[0x%llx,0x%llx) large_vma=[0x%llx,0x%llx) kernel_vma=[0x%llx,0x%llx)\n",
small_vma_start, small_vma_limit,
large_vma_start, large_vma_limit,
kernel_vma_start, kernel_vma_limit);
/* check that starts do not exceed limits */
WARN_ON(small_vma_start > small_vma_limit);
WARN_ON(large_vma_start > large_vma_limit);
/* kernel_vma must also be non-zero */
WARN_ON(kernel_vma_start >= kernel_vma_limit);
if (small_vma_start > small_vma_limit ||
large_vma_start > large_vma_limit ||
kernel_vma_start >= kernel_vma_limit) {
err = -EINVAL;
goto clean_up_pdes;
}
/*
* Attempt to make a separate VM for fixed allocations.
*/
if (g->separate_fixed_allocs &&
small_vma_start < small_vma_limit) {
if (g->separate_fixed_allocs >= small_vma_limit)
goto clean_up_pdes;
snprintf(alloc_name, sizeof(alloc_name),
"gk20a_%s-fixed", name);
err = __gk20a_allocator_init(&vm->fixed,
vm, alloc_name,
small_vma_start,
g->separate_fixed_allocs,
SZ_4K,
GPU_BALLOC_MAX_ORDER,
GPU_BALLOC_GVA_SPACE);
if (err)
goto clean_up_ptes;
/* Make sure to update the user vma size. */
small_vma_start = g->separate_fixed_allocs;
}
if (small_vma_start < small_vma_limit) {
snprintf(alloc_name, sizeof(alloc_name), "gk20a_%s-%dKB", name,
vm->gmmu_page_sizes[gmmu_page_size_small] >> 10);
err = __gk20a_allocator_init(&vm->vma[gmmu_page_size_small],
vm, alloc_name,
small_vma_start,
small_vma_limit - small_vma_start,
SZ_4K,
GPU_BALLOC_MAX_ORDER,
GPU_BALLOC_GVA_SPACE);
if (err)
goto clean_up_ptes;
}
if (large_vma_start < large_vma_limit) {
snprintf(alloc_name, sizeof(alloc_name), "gk20a_%s-%dKB",
name, vm->gmmu_page_sizes[gmmu_page_size_big] >> 10);
err = __gk20a_allocator_init(&vm->vma[gmmu_page_size_big],
vm, alloc_name,
large_vma_start,
large_vma_limit - large_vma_start,
big_page_size,
GPU_BALLOC_MAX_ORDER,
GPU_BALLOC_GVA_SPACE);
if (err)
goto clean_up_small_allocator;
}
snprintf(alloc_name, sizeof(alloc_name), "gk20a_%s-%dKB-sys",
name, vm->gmmu_page_sizes[gmmu_page_size_kernel] >> 10);
/*
* kernel reserved VMA is at the end of the aperture
*/
err = __gk20a_allocator_init(&vm->vma[gmmu_page_size_kernel],
vm, alloc_name,
kernel_vma_start,
kernel_vma_limit - kernel_vma_start,
SZ_4K,
GPU_BALLOC_MAX_ORDER,
GPU_BALLOC_GVA_SPACE);
if (err)
goto clean_up_big_allocator;
vm->mapped_buffers = RB_ROOT;
mutex_init(&vm->update_gmmu_lock);
kref_init(&vm->ref);
INIT_LIST_HEAD(&vm->reserved_va_list);
return 0;
clean_up_big_allocator:
if (large_vma_start < large_vma_limit)
gk20a_allocator_destroy(&vm->vma[gmmu_page_size_big]);
clean_up_small_allocator:
if (small_vma_start < small_vma_limit)
gk20a_allocator_destroy(&vm->vma[gmmu_page_size_small]);
clean_up_ptes:
free_gmmu_pages(vm, &vm->pdb);
clean_up_pdes:
vfree(vm->pdb.entries);
return err;
}
/* address space interfaces for the gk20a module */
int gk20a_vm_alloc_share(struct gk20a_as_share *as_share, u32 big_page_size,
u32 flags)
{
struct gk20a_as *as = as_share->as;
struct gk20a *g = gk20a_from_as(as);
struct mm_gk20a *mm = &g->mm;
struct vm_gk20a *vm;
char name[32];
int err;
const bool userspace_managed =
(flags & NVGPU_GPU_IOCTL_ALLOC_AS_FLAGS_USERSPACE_MANAGED) != 0;
gk20a_dbg_fn("");
if (big_page_size == 0) {
big_page_size =
gk20a_get_platform(g->dev)->default_big_page_size;
} else {
if (!is_power_of_2(big_page_size))
return -EINVAL;
if (!(big_page_size & g->gpu_characteristics.available_big_page_sizes))
return -EINVAL;
}
vm = kzalloc(sizeof(*vm), GFP_KERNEL);
if (!vm)
return -ENOMEM;
as_share->vm = vm;
vm->as_share = as_share;
vm->enable_ctag = true;
snprintf(name, sizeof(name), "gk20a_as_%d", as_share->id);
err = gk20a_init_vm(mm, vm, big_page_size, big_page_size << 10,
mm->channel.kernel_size,
mm->channel.user_size + mm->channel.kernel_size,
!mm->disable_bigpage, userspace_managed, name);
return err;
}
int gk20a_vm_release_share(struct gk20a_as_share *as_share)
{
struct vm_gk20a *vm = as_share->vm;
gk20a_dbg_fn("");
vm->as_share = NULL;
/* put as reference to vm */
gk20a_vm_put(vm);
as_share->vm = NULL;
return 0;
}
int gk20a_vm_alloc_space(struct gk20a_as_share *as_share,
struct nvgpu_as_alloc_space_args *args)
{
int err = -ENOMEM;
int pgsz_idx = gmmu_page_size_small;
struct gk20a_allocator *vma;
struct vm_gk20a *vm = as_share->vm;
struct gk20a *g = vm->mm->g;
struct vm_reserved_va_node *va_node;
u64 vaddr_start = 0;
int page_sizes = gmmu_nr_page_sizes;
gk20a_dbg_fn("flags=0x%x pgsz=0x%x nr_pages=0x%x o/a=0x%llx",
args->flags, args->page_size, args->pages,
args->o_a.offset);
if (!vm->big_pages)
page_sizes--;
for (; pgsz_idx < page_sizes; pgsz_idx++) {
if (vm->gmmu_page_sizes[pgsz_idx] == args->page_size)
break;
}
if (pgsz_idx >= page_sizes) {
err = -EINVAL;
goto clean_up;
}
va_node = kzalloc(sizeof(*va_node), GFP_KERNEL);
if (!va_node) {
err = -ENOMEM;
goto clean_up;
}
vma = &vm->vma[pgsz_idx];
if (args->flags & NVGPU_AS_ALLOC_SPACE_FLAGS_FIXED_OFFSET) {
if (vm->fixed.init)
vma = &vm->fixed;
vaddr_start = gk20a_balloc_fixed(vma, args->o_a.offset,
(u64)args->pages *
(u64)args->page_size);
} else {
vaddr_start = gk20a_balloc(vma,
(u64)args->pages *
(u64)args->page_size);
}
if (!vaddr_start) {
kfree(va_node);
goto clean_up;
}
va_node->vaddr_start = vaddr_start;
va_node->size = (u64)args->page_size * (u64)args->pages;
va_node->pgsz_idx = pgsz_idx;
INIT_LIST_HEAD(&va_node->va_buffers_list);
INIT_LIST_HEAD(&va_node->reserved_va_list);
mutex_lock(&vm->update_gmmu_lock);
/* mark that we need to use sparse mappings here */
if (args->flags & NVGPU_AS_ALLOC_SPACE_FLAGS_SPARSE) {
u64 map_offset = g->ops.mm.gmmu_map(vm, vaddr_start,
NULL,
0,
va_node->size,
pgsz_idx,
0,
0,
args->flags,
gk20a_mem_flag_none,
false,
true,
false,
NULL);
if (!map_offset) {
mutex_unlock(&vm->update_gmmu_lock);
gk20a_bfree(vma, vaddr_start);
kfree(va_node);
goto clean_up;
}
va_node->sparse = true;
}
list_add_tail(&va_node->reserved_va_list, &vm->reserved_va_list);
mutex_unlock(&vm->update_gmmu_lock);
args->o_a.offset = vaddr_start;
err = 0;
clean_up:
return err;
}
int gk20a_vm_free_space(struct gk20a_as_share *as_share,
struct nvgpu_as_free_space_args *args)
{
int err = -ENOMEM;
int pgsz_idx;
struct gk20a_allocator *vma;
struct vm_gk20a *vm = as_share->vm;
struct vm_reserved_va_node *va_node;
struct gk20a *g = gk20a_from_vm(vm);
gk20a_dbg_fn("pgsz=0x%x nr_pages=0x%x o/a=0x%llx", args->page_size,
args->pages, args->offset);
/* determine pagesz idx */
pgsz_idx = __nv_gmmu_va_is_big_page_region(vm, args->offset) ?
gmmu_page_size_big : gmmu_page_size_small;
if (vm->fixed.init)
vma = &vm->fixed;
else
vma = &vm->vma[pgsz_idx];
gk20a_bfree(vma, args->offset);
mutex_lock(&vm->update_gmmu_lock);
va_node = addr_to_reservation(vm, args->offset);
if (va_node) {
struct mapped_buffer_node *buffer, *n;
/* Decrement the ref count on all buffers in this va_node. This
* allows userspace to let the kernel free mappings that are
* only used by this va_node. */
list_for_each_entry_safe(buffer, n,
&va_node->va_buffers_list, va_buffers_list) {
list_del_init(&buffer->va_buffers_list);
kref_put(&buffer->ref, gk20a_vm_unmap_locked_kref);
}
list_del(&va_node->reserved_va_list);
/* if this was a sparse mapping, free the va */
if (va_node->sparse)
g->ops.mm.gmmu_unmap(vm,
va_node->vaddr_start,
va_node->size,
va_node->pgsz_idx,
true,
gk20a_mem_flag_none,
true,
NULL);
kfree(va_node);
}
mutex_unlock(&vm->update_gmmu_lock);
err = 0;
return err;
}
int gk20a_vm_bind_channel(struct gk20a_as_share *as_share,
struct channel_gk20a *ch)
{
int err = 0;
struct vm_gk20a *vm = as_share->vm;
gk20a_dbg_fn("");
ch->vm = vm;
err = channel_gk20a_commit_va(ch);
if (err)
ch->vm = NULL;
return err;
}
int gk20a_dmabuf_alloc_drvdata(struct dma_buf *dmabuf, struct device *dev)
{
struct gk20a_dmabuf_priv *priv;
static DEFINE_MUTEX(priv_lock);
static u64 priv_count = 0;
priv = dma_buf_get_drvdata(dmabuf, dev);
if (likely(priv))
return 0;
mutex_lock(&priv_lock);
priv = dma_buf_get_drvdata(dmabuf, dev);
if (priv)
goto priv_exist_or_err;
priv = kzalloc(sizeof(*priv), GFP_KERNEL);
if (!priv) {
priv = ERR_PTR(-ENOMEM);
goto priv_exist_or_err;
}
mutex_init(&priv->lock);
INIT_LIST_HEAD(&priv->states);
priv->buffer_id = ++priv_count;
dma_buf_set_drvdata(dmabuf, dev, priv, gk20a_mm_delete_priv);
priv_exist_or_err:
mutex_unlock(&priv_lock);
if (IS_ERR(priv))
return -ENOMEM;
return 0;
}
int gk20a_dmabuf_get_state(struct dma_buf *dmabuf, struct device *dev,
u64 offset, struct gk20a_buffer_state **state)
{
int err = 0;
struct gk20a_dmabuf_priv *priv;
struct gk20a_buffer_state *s;
if (WARN_ON(offset >= (u64)dmabuf->size))
return -EINVAL;
err = gk20a_dmabuf_alloc_drvdata(dmabuf, dev);
if (err)
return err;
priv = dma_buf_get_drvdata(dmabuf, dev);
if (WARN_ON(!priv))
return -ENOSYS;
mutex_lock(&priv->lock);
list_for_each_entry(s, &priv->states, list)
if (s->offset == offset)
goto out;
/* State not found, create state. */
s = kzalloc(sizeof(*s), GFP_KERNEL);
if (!s) {
err = -ENOMEM;
goto out;
}
s->offset = offset;
INIT_LIST_HEAD(&s->list);
mutex_init(&s->lock);
list_add_tail(&s->list, &priv->states);
out:
mutex_unlock(&priv->lock);
if (!err)
*state = s;
return err;
}
int gk20a_vm_map_buffer(struct vm_gk20a *vm,
int dmabuf_fd,
u64 *offset_align,
u32 flags, /*NVGPU_AS_MAP_BUFFER_FLAGS_*/
int kind,
u64 buffer_offset,
u64 mapping_size,
struct vm_gk20a_mapping_batch *batch)
{
int err = 0;
struct dma_buf *dmabuf;
u64 ret_va;
gk20a_dbg_fn("");
/* get ref to the mem handle (released on unmap_locked) */
dmabuf = dma_buf_get(dmabuf_fd);
if (IS_ERR(dmabuf)) {
dev_warn(dev_from_vm(vm), "%s: fd %d is not a dmabuf",
__func__, dmabuf_fd);
return PTR_ERR(dmabuf);
}
err = gk20a_dmabuf_alloc_drvdata(dmabuf, dev_from_vm(vm));
if (err) {
dma_buf_put(dmabuf);
return err;
}
ret_va = gk20a_vm_map(vm, dmabuf, *offset_align,
flags, kind, NULL, true,
gk20a_mem_flag_none,
buffer_offset,
mapping_size,
batch);
*offset_align = ret_va;
if (!ret_va) {
dma_buf_put(dmabuf);
err = -EINVAL;
}
return err;
}
int gk20a_vm_unmap_buffer(struct vm_gk20a *vm, u64 offset,
struct vm_gk20a_mapping_batch *batch)
{
gk20a_dbg_fn("");
gk20a_vm_unmap_user(vm, offset, batch);
return 0;
}
void gk20a_deinit_vm(struct vm_gk20a *vm)
{
gk20a_allocator_destroy(&vm->vma[gmmu_page_size_kernel]);
if (vm->vma[gmmu_page_size_big].init)
gk20a_allocator_destroy(&vm->vma[gmmu_page_size_big]);
if (vm->vma[gmmu_page_size_small].init)
gk20a_allocator_destroy(&vm->vma[gmmu_page_size_small]);
if (vm->fixed.init)
gk20a_allocator_destroy(&vm->fixed);
gk20a_vm_free_entries(vm, &vm->pdb, 0);
}
int gk20a_alloc_inst_block(struct gk20a *g, struct mem_desc *inst_block)
{
struct device *dev = dev_from_gk20a(g);
int err;
gk20a_dbg_fn("");
err = gk20a_gmmu_alloc(g, ram_in_alloc_size_v(), inst_block);
if (err) {
gk20a_err(dev, "%s: memory allocation failed\n", __func__);
return err;
}
gk20a_dbg_fn("done");
return 0;
}
void gk20a_free_inst_block(struct gk20a *g, struct mem_desc *inst_block)
{
if (inst_block->cpu_va)
gk20a_gmmu_free(g, inst_block);
}
u64 gk20a_mm_inst_block_addr(struct gk20a *g, struct mem_desc *inst_block)
{
u64 addr;
if (g->mm.has_physical_mode)
addr = gk20a_mem_phys(inst_block);
else
addr = gk20a_mm_smmu_vaddr_translate(g, sg_dma_address(inst_block->sgt->sgl));
return addr;
}
static int gk20a_init_bar1_vm(struct mm_gk20a *mm)
{
int err;
struct vm_gk20a *vm = &mm->bar1.vm;
struct gk20a *g = gk20a_from_mm(mm);
struct mem_desc *inst_block = &mm->bar1.inst_block;
u32 big_page_size = gk20a_get_platform(g->dev)->default_big_page_size;
mm->bar1.aperture_size = bar1_aperture_size_mb_gk20a() << 20;
gk20a_dbg_info("bar1 vm size = 0x%x", mm->bar1.aperture_size);
gk20a_init_vm(mm, vm, big_page_size, SZ_4K,
mm->bar1.aperture_size - SZ_4K,
mm->bar1.aperture_size, false, false, "bar1");
err = gk20a_alloc_inst_block(g, inst_block);
if (err)
goto clean_up_va;
gk20a_init_inst_block(inst_block, vm, big_page_size);
return 0;
clean_up_va:
gk20a_deinit_vm(vm);
return err;
}
/* pmu vm, share channel_vm interfaces */
static int gk20a_init_system_vm(struct mm_gk20a *mm)
{
int err;
struct vm_gk20a *vm = &mm->pmu.vm;
struct gk20a *g = gk20a_from_mm(mm);
struct mem_desc *inst_block = &mm->pmu.inst_block;
u32 big_page_size = gk20a_get_platform(g->dev)->default_big_page_size;
mm->pmu.aperture_size = GK20A_PMU_VA_SIZE;
gk20a_dbg_info("pmu vm size = 0x%x", mm->pmu.aperture_size);
gk20a_init_vm(mm, vm, big_page_size,
SZ_4K * 16, GK20A_PMU_VA_SIZE,
GK20A_PMU_VA_SIZE * 2, false, false,
"system");
err = gk20a_alloc_inst_block(g, inst_block);
if (err)
goto clean_up_va;
gk20a_init_inst_block(inst_block, vm, big_page_size);
return 0;
clean_up_va:
gk20a_deinit_vm(vm);
return err;
}
static int gk20a_init_hwpm(struct mm_gk20a *mm)
{
int err;
struct vm_gk20a *vm = &mm->pmu.vm;
struct gk20a *g = gk20a_from_mm(mm);
struct mem_desc *inst_block = &mm->hwpm.inst_block;
err = gk20a_alloc_inst_block(g, inst_block);
if (err)
return err;
gk20a_init_inst_block(inst_block, vm, 0);
return 0;
}
static int gk20a_init_cde_vm(struct mm_gk20a *mm)
{
struct vm_gk20a *vm = &mm->cde.vm;
struct gk20a *g = gk20a_from_mm(mm);
u32 big_page_size = gk20a_get_platform(g->dev)->default_big_page_size;
return gk20a_init_vm(mm, vm, big_page_size,
SZ_4K * 16,
NV_MM_DEFAULT_KERNEL_SIZE,
NV_MM_DEFAULT_KERNEL_SIZE + NV_MM_DEFAULT_USER_SIZE,
false, false, "cde");
}
void gk20a_mm_init_pdb(struct gk20a *g, struct mem_desc *mem, u64 pdb_addr)
{
u32 pdb_addr_lo = u64_lo32(pdb_addr >> ram_in_base_shift_v());
u32 pdb_addr_hi = u64_hi32(pdb_addr);
gk20a_mem_wr32(g, mem, ram_in_page_dir_base_lo_w(),
(g->mm.vidmem_is_vidmem ?
ram_in_page_dir_base_target_sys_mem_ncoh_f() :
ram_in_page_dir_base_target_vid_mem_f()) |
ram_in_page_dir_base_vol_true_f() |
ram_in_page_dir_base_lo_f(pdb_addr_lo));
gk20a_mem_wr32(g, mem, ram_in_page_dir_base_hi_w(),
ram_in_page_dir_base_hi_f(pdb_addr_hi));
}
void gk20a_init_inst_block(struct mem_desc *inst_block, struct vm_gk20a *vm,
u32 big_page_size)
{
struct gk20a *g = gk20a_from_vm(vm);
u64 pde_addr = g->ops.mm.get_iova_addr(g, vm->pdb.mem.sgt->sgl, 0);
gk20a_dbg_info("inst block phys = 0x%llx, kv = 0x%p",
gk20a_mm_inst_block_addr(g, inst_block), inst_block->cpu_va);
gk20a_dbg_info("pde pa=0x%llx", (u64)pde_addr);
g->ops.mm.init_pdb(g, inst_block, pde_addr);
gk20a_mem_wr32(g, inst_block, ram_in_adr_limit_lo_w(),
u64_lo32(vm->va_limit - 1) & ~0xfff);
gk20a_mem_wr32(g, inst_block, ram_in_adr_limit_hi_w(),
ram_in_adr_limit_hi_f(u64_hi32(vm->va_limit - 1)));
if (big_page_size && g->ops.mm.set_big_page_size)
g->ops.mm.set_big_page_size(g, inst_block, big_page_size);
}
int gk20a_mm_fb_flush(struct gk20a *g)
{
struct mm_gk20a *mm = &g->mm;
u32 data;
s32 retry = 100;
int ret = 0;
gk20a_dbg_fn("");
gk20a_busy_noresume(g->dev);
if (!g->power_on) {
pm_runtime_put_noidle(g->dev);
return 0;
}
mutex_lock(&mm->l2_op_lock);
/* Make sure all previous writes are committed to the L2. There's no
guarantee that writes are to DRAM. This will be a sysmembar internal
to the L2. */
trace_gk20a_mm_fb_flush(dev_name(g->dev));
gk20a_writel(g, flush_fb_flush_r(),
flush_fb_flush_pending_busy_f());
do {
data = gk20a_readl(g, flush_fb_flush_r());
if (flush_fb_flush_outstanding_v(data) ==
flush_fb_flush_outstanding_true_v() ||
flush_fb_flush_pending_v(data) ==
flush_fb_flush_pending_busy_v()) {
gk20a_dbg_info("fb_flush 0x%x", data);
retry--;
udelay(5);
} else
break;
} while (retry >= 0 || !tegra_platform_is_silicon());
if (tegra_platform_is_silicon() && retry < 0) {
gk20a_warn(dev_from_gk20a(g),
"fb_flush too many retries");
if (g->ops.fb.dump_vpr_wpr_info)
g->ops.fb.dump_vpr_wpr_info(g);
ret = -EBUSY;
}
trace_gk20a_mm_fb_flush_done(dev_name(g->dev));
mutex_unlock(&mm->l2_op_lock);
pm_runtime_put_noidle(g->dev);
return ret;
}
static void gk20a_mm_l2_invalidate_locked(struct gk20a *g)
{
u32 data;
s32 retry = 200;
trace_gk20a_mm_l2_invalidate(dev_name(g->dev));
/* Invalidate any clean lines from the L2 so subsequent reads go to
DRAM. Dirty lines are not affected by this operation. */
gk20a_writel(g, flush_l2_system_invalidate_r(),
flush_l2_system_invalidate_pending_busy_f());
do {
data = gk20a_readl(g, flush_l2_system_invalidate_r());
if (flush_l2_system_invalidate_outstanding_v(data) ==
flush_l2_system_invalidate_outstanding_true_v() ||
flush_l2_system_invalidate_pending_v(data) ==
flush_l2_system_invalidate_pending_busy_v()) {
gk20a_dbg_info("l2_system_invalidate 0x%x",
data);
retry--;
udelay(5);
} else
break;
} while (retry >= 0 || !tegra_platform_is_silicon());
if (tegra_platform_is_silicon() && retry < 0)
gk20a_warn(dev_from_gk20a(g),
"l2_system_invalidate too many retries");
trace_gk20a_mm_l2_invalidate_done(dev_name(g->dev));
}
void gk20a_mm_l2_invalidate(struct gk20a *g)
{
struct mm_gk20a *mm = &g->mm;
gk20a_busy_noresume(g->dev);
if (g->power_on) {
mutex_lock(&mm->l2_op_lock);
gk20a_mm_l2_invalidate_locked(g);
mutex_unlock(&mm->l2_op_lock);
}
pm_runtime_put_noidle(g->dev);
}
void gk20a_mm_l2_flush(struct gk20a *g, bool invalidate)
{
struct mm_gk20a *mm = &g->mm;
u32 data;
s32 retry = 200;
gk20a_dbg_fn("");
gk20a_busy_noresume(g->dev);
if (!g->power_on)
goto hw_was_off;
mutex_lock(&mm->l2_op_lock);
trace_gk20a_mm_l2_flush(dev_name(g->dev));
/* Flush all dirty lines from the L2 to DRAM. Lines are left in the L2
as clean, so subsequent reads might hit in the L2. */
gk20a_writel(g, flush_l2_flush_dirty_r(),
flush_l2_flush_dirty_pending_busy_f());
do {
data = gk20a_readl(g, flush_l2_flush_dirty_r());
if (flush_l2_flush_dirty_outstanding_v(data) ==
flush_l2_flush_dirty_outstanding_true_v() ||
flush_l2_flush_dirty_pending_v(data) ==
flush_l2_flush_dirty_pending_busy_v()) {
gk20a_dbg_info("l2_flush_dirty 0x%x", data);
retry--;
udelay(5);
} else
break;
} while (retry >= 0 || !tegra_platform_is_silicon());
if (tegra_platform_is_silicon() && retry < 0)
gk20a_warn(dev_from_gk20a(g),
"l2_flush_dirty too many retries");
trace_gk20a_mm_l2_flush_done(dev_name(g->dev));
if (invalidate)
gk20a_mm_l2_invalidate_locked(g);
mutex_unlock(&mm->l2_op_lock);
hw_was_off:
pm_runtime_put_noidle(g->dev);
}
void gk20a_mm_cbc_clean(struct gk20a *g)
{
struct mm_gk20a *mm = &g->mm;
u32 data;
s32 retry = 200;
gk20a_dbg_fn("");
gk20a_busy_noresume(g->dev);
if (!g->power_on)
goto hw_was_off;
mutex_lock(&mm->l2_op_lock);
/* Flush all dirty lines from the CBC to L2 */
gk20a_writel(g, flush_l2_clean_comptags_r(),
flush_l2_clean_comptags_pending_busy_f());
do {
data = gk20a_readl(g, flush_l2_clean_comptags_r());
if (flush_l2_clean_comptags_outstanding_v(data) ==
flush_l2_clean_comptags_outstanding_true_v() ||
flush_l2_clean_comptags_pending_v(data) ==
flush_l2_clean_comptags_pending_busy_v()) {
gk20a_dbg_info("l2_clean_comptags 0x%x", data);
retry--;
udelay(5);
} else
break;
} while (retry >= 0 || !tegra_platform_is_silicon());
if (tegra_platform_is_silicon() && retry < 0)
gk20a_warn(dev_from_gk20a(g),
"l2_clean_comptags too many retries");
mutex_unlock(&mm->l2_op_lock);
hw_was_off:
pm_runtime_put_noidle(g->dev);
}
int gk20a_vm_find_buffer(struct vm_gk20a *vm, u64 gpu_va,
struct dma_buf **dmabuf,
u64 *offset)
{
struct mapped_buffer_node *mapped_buffer;
gk20a_dbg_fn("gpu_va=0x%llx", gpu_va);
mutex_lock(&vm->update_gmmu_lock);
mapped_buffer = find_mapped_buffer_range_locked(&vm->mapped_buffers,
gpu_va);
if (!mapped_buffer) {
mutex_unlock(&vm->update_gmmu_lock);
return -EINVAL;
}
*dmabuf = mapped_buffer->dmabuf;
*offset = gpu_va - mapped_buffer->addr;
mutex_unlock(&vm->update_gmmu_lock);
return 0;
}
void gk20a_mm_tlb_invalidate(struct vm_gk20a *vm)
{
struct gk20a *g = gk20a_from_vm(vm);
u32 addr_lo = u64_lo32(g->ops.mm.get_iova_addr(vm->mm->g,
vm->pdb.mem.sgt->sgl, 0) >> 12);
u32 data;
s32 retry = 2000;
static DEFINE_MUTEX(tlb_lock);
gk20a_dbg_fn("");
/* pagetables are considered sw states which are preserved after
prepare_poweroff. When gk20a deinit releases those pagetables,
common code in vm unmap path calls tlb invalidate that touches
hw. Use the power_on flag to skip tlb invalidation when gpu
power is turned off */
if (!g->power_on)
return;
mutex_lock(&tlb_lock);
trace_gk20a_mm_tlb_invalidate(dev_name(g->dev));
do {
data = gk20a_readl(g, fb_mmu_ctrl_r());
if (fb_mmu_ctrl_pri_fifo_space_v(data) != 0)
break;
udelay(2);
retry--;
} while (retry >= 0 || !tegra_platform_is_silicon());
if (tegra_platform_is_silicon() && retry < 0) {
gk20a_warn(dev_from_gk20a(g),
"wait mmu fifo space too many retries");
goto out;
}
gk20a_writel(g, fb_mmu_invalidate_pdb_r(),
fb_mmu_invalidate_pdb_addr_f(addr_lo) |
(g->mm.vidmem_is_vidmem ?
fb_mmu_invalidate_pdb_aperture_sys_mem_f() :
fb_mmu_invalidate_pdb_aperture_vid_mem_f()));
gk20a_writel(g, fb_mmu_invalidate_r(),
fb_mmu_invalidate_all_va_true_f() |
fb_mmu_invalidate_trigger_true_f());
do {
data = gk20a_readl(g, fb_mmu_ctrl_r());
if (fb_mmu_ctrl_pri_fifo_empty_v(data) !=
fb_mmu_ctrl_pri_fifo_empty_false_f())
break;
retry--;
udelay(2);
} while (retry >= 0 || !tegra_platform_is_silicon());
if (tegra_platform_is_silicon() && retry < 0)
gk20a_warn(dev_from_gk20a(g),
"mmu invalidate too many retries");
trace_gk20a_mm_tlb_invalidate_done(dev_name(g->dev));
out:
mutex_unlock(&tlb_lock);
}
int gk20a_mm_suspend(struct gk20a *g)
{
gk20a_dbg_fn("");
g->ops.mm.cbc_clean(g);
g->ops.mm.l2_flush(g, false);
gk20a_dbg_fn("done");
return 0;
}
bool gk20a_mm_mmu_debug_mode_enabled(struct gk20a *g)
{
u32 debug_ctrl = gk20a_readl(g, fb_mmu_debug_ctrl_r());
return fb_mmu_debug_ctrl_debug_v(debug_ctrl) ==
fb_mmu_debug_ctrl_debug_enabled_v();
}
static void gk20a_mm_mmu_set_debug_mode(struct gk20a *g, bool enable)
{
u32 reg_val, debug_ctrl;
reg_val = gk20a_readl(g, fb_mmu_debug_ctrl_r());
if (enable) {
debug_ctrl = fb_mmu_debug_ctrl_debug_enabled_f();
g->mmu_debug_ctrl = true;
} else {
debug_ctrl = fb_mmu_debug_ctrl_debug_disabled_f();
g->mmu_debug_ctrl = false;
}
reg_val = set_field(reg_val,
fb_mmu_debug_ctrl_debug_m(), debug_ctrl);
gk20a_writel(g, fb_mmu_debug_ctrl_r(), reg_val);
}
u32 gk20a_mm_get_physical_addr_bits(struct gk20a *g)
{
return 34;
}
const struct gk20a_mmu_level *gk20a_mm_get_mmu_levels(struct gk20a *g,
u32 big_page_size)
{
return (big_page_size == SZ_64K) ?
gk20a_mm_levels_64k : gk20a_mm_levels_128k;
}
int gk20a_mm_get_buffer_info(struct device *dev, int dmabuf_fd,
u64 *buffer_id, u64 *buffer_len)
{
struct dma_buf *dmabuf;
struct gk20a_dmabuf_priv *priv;
int err = 0;
dmabuf = dma_buf_get(dmabuf_fd);
if (IS_ERR(dmabuf)) {
dev_warn(dev, "%s: fd %d is not a dmabuf", __func__, dmabuf_fd);
return PTR_ERR(dmabuf);
}
err = gk20a_dmabuf_alloc_drvdata(dmabuf, dev);
if (err) {
dev_warn(dev, "Failed to allocate dmabuf drvdata (err = %d)",
err);
goto clean_up;
}
priv = dma_buf_get_drvdata(dmabuf, dev);
if (likely(priv)) {
*buffer_id = priv->buffer_id;
*buffer_len = dmabuf->size;
}
clean_up:
dma_buf_put(dmabuf);
return err;
}
void gk20a_mm_debugfs_init(struct device *dev)
{
struct gk20a_platform *platform = dev_get_drvdata(dev);
struct dentry *gpu_root = platform->debugfs;
struct gk20a *g = gk20a_get_platform(dev)->g;
debugfs_create_x64("separate_fixed_allocs", 0664, gpu_root,
&g->separate_fixed_allocs);
}
void gk20a_init_mm(struct gpu_ops *gops)
{
gops->mm.is_debug_mode_enabled = gk20a_mm_mmu_debug_mode_enabled;
gops->mm.set_debug_mode = gk20a_mm_mmu_set_debug_mode;
gops->mm.gmmu_map = gk20a_locked_gmmu_map;
gops->mm.gmmu_unmap = gk20a_locked_gmmu_unmap;
gops->mm.vm_remove = gk20a_vm_remove_support;
gops->mm.vm_alloc_share = gk20a_vm_alloc_share;
gops->mm.vm_bind_channel = gk20a_vm_bind_channel;
gops->mm.fb_flush = gk20a_mm_fb_flush;
gops->mm.l2_invalidate = gk20a_mm_l2_invalidate;
gops->mm.l2_flush = gk20a_mm_l2_flush;
gops->mm.cbc_clean = gk20a_mm_cbc_clean;
gops->mm.tlb_invalidate = gk20a_mm_tlb_invalidate;
gops->mm.get_iova_addr = gk20a_mm_iova_addr;
gops->mm.get_physical_addr_bits = gk20a_mm_get_physical_addr_bits;
gops->mm.get_mmu_levels = gk20a_mm_get_mmu_levels;
gops->mm.init_pdb = gk20a_mm_init_pdb;
gops->mm.init_mm_setup_hw = gk20a_init_mm_setup_hw;
gops->mm.bar1_bind = gk20a_mm_bar1_bind;
}
|